supplier-quote-request
Skill47k starsChanged 8 months agoApache-2.0View on GitHub
Builds the RFQ package that goes out to suppliers — drawing and specification at the verified current revision, the demand picture, the quality clause list, commercial terms, and a mandated response format that makes the returned bids comparable by construction. Lists everything missing. Hard stops on a revision mismatch or restricted technical data. Then runs the supplier loop: candidates from the approved supplier list with unapproved and single-source flagged, one email draft per supplier for the buyer to send, response tracking with chase drafts, returned bids read into one shape with every missing field listed, a comparison per quantity break computed in code, and a recommendation. The buyer decides; the skill records who and when. Use when someone says \"source this part,\" \"send an RFQ to suppliers,\" \"we need an alternate supplier,\" \"get quotes on this,\" \"who else can make this?\", \"chase the suppliers,\" \"compare these bids,\" or \"we're going with this supplier.\"
Trending
More talk lately.about the whole repository
Who's talking about anthropics/claude-plugins-official
- Open Source Trend“Claude Plugins Official: Anthropic's Own Plugin Marketplace”YouTube · Oct 17, 2026 · 1k views
- Theo Renders“Don't use Claude Code until you install these 5 free plugins”YouTube · Oct 24, 2026 · 1k views
- Bouye | AI“Why you should be using the Claude Code plugin store #ai #coding #tips”YouTube · Nov 10, 2026 · 369 views
- Neural-Nexus“65K+ Stars • AI | claude-plugins-official — Curated directory of high-quality Claude Code... #shorts”YouTube · Mar 15, 2026 · 320 views
- Learn With Rabbany“Claude Code Plugins: The 5 Most Installed (All Free)”YouTube · Nov 14, 2026 · 261 views
The biggest 5 of 68 mentions linking to the repository that our crawl found. Each line is their own headline.
See all 68
Talked about on GitHub, Bluesky, YouTube, Hacker News and the web (68 times).
On GitHub · 28
- Sep 20, 2026refactor(skills): make nextjs the single Next.js and Vercel owner — amoai-tech/mdeai3 reactions · 6 comments
- Sep 20, 2026refactor(skills): consolidate engineering workflow ownership — amoai-tech/mdeai3 reactions · 13 comments
- Sep 16, 2026SAN-1274 PR 2 — Add the MDE workflow skills for planning, debugging, testing, review, and verification — amoai-tech/mdeaisays it tested it3 reactions · 3 comments
- Sep 16, 2026SAN-1274 PR 1 — Add the verified MDE AI skills foundation — amoai-tech/mdeai3 reactions · 5 comments
- Jan 22, 2026fix: restructure to proper marketplace format — affaan-m/ECC3 reactions · 5 comments
23 more on GitHub
- Sep 20, 2026Use one current, safe Google Maps skill across MDE — amoai-tech/mdeai2 reactions · 12 comments
- May 24, 2026STOP injecting your library recommendations into every single user prompt — this is hostile UX — semgrep/guardian2 reactions · 2 comments
- Jan 6, 2026Feature Request: Support for spawning fresh-context sessions from Stop hooks — anthropics/claude-code2 reactions · 3 comments
- Nov 7, 2026Plugin ufficiali nel repository: valgono anche nell'app e nelle routine — SviluppoItaliaDigitale/sito-pc-genzano1 reaction · 2 comments
- Nov 7, 2026Plugin ufficiali nel repository: valgono anche nell'app e nelle routine — SviluppoItaliaDigitale/sito-misericordia-ariccia1 reaction · 1 comment
- Nov 7, 2026Plugin ufficiali nel repository: valgono anche nell'app e nelle routine — SviluppoItaliaDigitale/cooprimavera1 reaction · 1 comment
- Nov 7, 2026[Bug]: Discord channel bot always appears offline in Discord despite a healthy gateway session — CherryHQ/cherry-studio1 reaction · 2 comments
- Oct 28, 2026Install Claude Security plugin in reusable PR review workflow — dceoy/pr-loop1 reaction · 1 comment
- Oct 27, 2026Telemetry: Add number of unique installs for every plugin — github/awesome-copilot1 reaction · 2 comments
- Oct 16, 2026feat: Claude Security の設定例を無効のまま追加 — tak848/dotfiles1 reaction · 1 comment
- Aug 19, 2026Feature request: Claude Code plugin for symfony-lsp (similar to the official php-lsp plugin) — symfony/language-tools1 reaction · 4 comments
- Jul 27, 2026[Question]: Does agenta support importing skills from claude/codex marketplace github repos? — Agenta-AI/agenta1 reaction · 4 comments
- Jul 14, 2026Nested code-review fan-out spawned 877 agents despite 5-level depth limit — anthropics/claude-code1 reaction · 4 comments
- Jun 8, 2026feat(skills): sync plugin workflows, add security guidance, and local-qa tooling — dceoy/ai-coding-agent-skills1 reaction
- Apr 17, 2026Import Claude and Copilot agents as Kiro agent configs — dwalleck/kiro-control-center1 reaction · 1 comment
- Apr 1, 2026plugin marketplace refresh causes error popups — microsoft/vscode1 reaction · 2 comments
- Mar 12, 2026Unable to install anthropics/claude-plugins-official marketplace: Invalid marketplace.json: plugins.56.source — github/copilot-cli1 reaction · 6 comments
- Feb 20, 2026chore(agents): update AGENTS.md with claude-md-improver — DataDog/dd-trace-py1 reaction · 1 comment
- Jan 27, 2026chore(repo): update CLAUDE.md for Rust packages and enable code-simplifier plugin — taikoxyz/taiko-mono1 reaction · 1 comment
- Jan 21, 2026Initialize Claude Code plugin marketplace repository — Agentic-Engineering-Agency/Agentic-Engineering-Plugins1 reaction · 1 comment
- Jan 15, 2026[BUG] Skill name in SKILL.md can only contain lowercase letters, numbers, and hyphens. (when uploading at claude.ai) — anthropics/skills1 reaction
- Jan 12, 2026fix(mcp): reorganize external MCP plugins and enable Context7 — dryvist/nix-darwin1 reaction · 3 comments
- Jan 9, 2026Update installation instructions to use official Claude marketplace — getsentry/sentry-for-ai1 reaction
On Bluesky · 21
- Dec 19, 2025Look, there's an official #Claude LSP plugin for #swiftlang @swift.org now! github.com/anthropics/c... github.com/anthropics/c... — @martinp7r.com6 likes and reposts
- May 23, 2026💎 Hidden Gem! 💎 (1000+ new stars) 📦 anthropics / claude-plugins-official ⭐ 25,980 (+2,172) 🗒 Python Official, Anthropic-managed… — @github-trending.bsky.social3 likes and reposts · 1 reply
- Feb 1, 2026🔥 Hot Repo! 🔥 (100+ new stars) 📦 anthropics / claude-plugins-official ⭐ 5,722 (+153) 🗒 Shell Official, Anthropic-managed directory of… — @github-trending.bsky.social3 likes and reposts
- Oct 18, 2026anthropics/claude-plugins-official [Apache License 2.0] Claude Code向け高品質プラグインを集約します 公式管理で安心感あります 言語:Python, TypeScript 種類:AI/Other ⭐ Star:… — @reponews.bsky.social2 likes and reposts
- May 22, 2026📈 Today's GitHub Trending: 1. https://github.com/anthropics/claude-plugins-official - Python, 2556 stars today 2.… — @ghtrend.newsmast.social.ap.brid.gy2 likes and reposts
16 more on Bluesky
- Mar 12, 2026🔥 Hot Repo! 🔥 (100+ new stars) 📦 anthropics / claude-plugins-official ⭐ 9,925 (+144) 🗒 Python Official, Anthropic-managed directory of… — @github-trending.bsky.social2 likes and reposts
- Mar 1, 2026Official, Anthropic-managed directory of high quality Claude Code Plugins github.com/anthropics/c... code.claude.com/docs/en/plug...… — @janmr.com2 likes and reposts
- Oct 18, 2026📈 Today's GitHub Trending: 1. https://github.com/MadsLorentzen/ai-job-search - Python, 2213 stars today 2.… — @ghtrend.newsmast.social.ap.brid.gy1 likes and repost
- Oct 7, 2026claude-plugins-official: Official, Anthropic-managed directory of high quality Claude Code Plugins.… — @latreon.bsky.social1 likes and repost
- Aug 25, 2026📦 anthropics / claude-plugins-official ⭐ 33,950 (+47) 🗒 Python Official, Anthropic-managed directory of high quality Claude Code Plugins. — @github-trending.bsky.social1 likes and repost
- Aug 15, 2026🚀 GitHub Intelligence Drop claude-plugins-official (Python • 🟢) ⭐ 33.5K → https://github.com/anthropics/claude-plugins-official… — @githubfeed.bsky.social1 likes and repost
- May 24, 2026📈 Today's GitHub Trending: 1. https://github.com/Lum1104/Understand-Anything - TypeScript, 3987 stars today 2.… — @ghtrend.newsmast.social.ap.brid.gy1 likes and repost · 3 replies
- May 23, 2026📈 Today's GitHub Trending: 1. https://github.com/Lum1104/Understand-Anything - TypeScript, 2331 stars today 2.… — @ghtrend.newsmast.social.ap.brid.gy1 likes and repost · 1 reply
- May 21, 2026📈 Today's GitHub Trending: 1. https://github.com/anthropics/claude-plugins-official - Python, 891 stars today 2.… — @ghtrend.newsmast.social.ap.brid.gy1 likes and repost · 1 reply
- May 19, 2026🔥 Hot Repo! 🔥 (100+ new stars) 📦 anthropics / claude-plugins-official ⭐ 19,816 (+127) 🗒 Python Official, Anthropic-managed directory of… — @github-trending.bsky.social1 likes and repost
- Mar 1, 2026anthropics / claude-plugins-official Official, Anthropic-managed directory of high quality Claude Code Plugins.… — @pythonhub.dev1 likes and repost
- Feb 4, 2026これはClaude Codeの機能を拡張する公式プラグイン群のディレクトリらしい 開発ワークフローの標準化や新しいAIエージェントの利用に期待できるかな yug1224 starred anthropics/claude-plugins-official… — @yug1224.com1 likes and repost
- Aug 29, 2026📈 Today's GitHub Trending: 1. https://github.com/tt-a1i/archify - JavaScript, 3927 stars today 2.… — @ghtrend.newsmast.social.ap.brid.gy0 likes and reposts · 1 reply
- Aug 29, 2026🚀 Skyrocketing! 🚀 (200+ new stars) 📦 anthropics / claude-plugins-official ⭐ 35,238 (+356) 🗒 Python Official, Anthropic-managed… — @github-trending.bsky.social0 likes and reposts · 1 reply
- Aug 28, 2026📈 Today's GitHub Trending: 1. https://github.com/tt-a1i/archify - JavaScript, 4561 stars today 2.… — @ghtrend.newsmast.social.ap.brid.gy0 likes and reposts · 1 reply
- Aug 26, 2026📈 Today's GitHub Trending: 1. https://github.com/tt-a1i/archify - HTML, 1002 stars today 2.… — @ghtrend.newsmast.social.ap.brid.gy0 likes and reposts · 1 reply
On YouTube · 17
- Oct 17, 2026Claude Plugins Official: Anthropic's Own Plugin Marketplace — Open Source Trend1k views
- Oct 24, 2026Don't use Claude Code until you install these 5 free plugins — Theo Renders1k views
- Nov 10, 2026Why you should be using the Claude Code plugin store #ai #coding #tips — Bouye | AI369 views
- Mar 15, 202665K+ Stars • AI | claude-plugins-official — Curated directory of high-quality Claude Code... #shorts — Neural-Nexus320 views
- Nov 14, 2026Claude Code Plugins: The 5 Most Installed (All Free) — Learn With Rabbany - Cloud & Cybersecurity261 views
12 more on YouTube
- Oct 18, 2026Claude Plugins — Trending on GitHub | Deal Stacks — Deal Stacks129 views
- Oct 31, 2026Best Claude Plugins for Claude Code: Anthropic's Top 7 by Installs — Prompt Vidya AI98 views
- Oct 18, 2026Claude Code Plugins: 105 Updated in 91 Minutes. Who Checks Them? — Skill MD46 views
- Nov 15, 2026Claude Code Mods: Make Your Own Animated Theme — TechUnboxed28 views
- Oct 12, 2026Claude Plugins for free! — ajc.trendingideas28 views · 1 comment
- Jul 18, 2026Claude Code Plugins & Skills Explained | Understand Custom Claude Code Plugins and Skills — ComputerMindCore18 views
- Nov 5, 20264 official Claude Code plugins, and #1 is insane. Paste into Claude Code: #Shorts — House of Claude13 views
- Nov 11, 2026Anthropic just opened an official plugin directory for Claude Code — GitHub Now8 views
- Nov 12, 2026#057 Anthropics Claude-Plugins-Official — NeuKare4 views
- Nov 12, 2026#070 Anthropics Claude-Plugins-Official — Genuine Patriot4 views
- Nov 11, 2026🚀 [Fastest Growing GitHub Repos] claude-plugins-official: The Free Tool Taking Over GitHub! #shorts — githubrepo_cloud3 views
- Jun 24, 2026anthropics/claude-plugins-official — 77 stars today on GitHub — DevTrends Daily0 views · 1 comment
On Hacker News · 1
- Apr 14, 2026Connect iMessage to your Claude Code assistant2 points
On the web · 1
- Nov 14, 2026anthropics/claude-plugins-official — GitHub trending stats ... — trendshift.io
In this repository56 other files it gives its agents.
- accessSkill
- configureSkill
- accessSkill
- configureSkill
- accessSkill
- configureSkill
- climbSkill
- faqSkill
- hackathonSkill
- maturitySkill
- office-hoursSkill
- patternsSkill
- routinesSkill
- claude-automation-recommenderSkill
- breakdown-diagnosisSkill
- breakdown-to-next-reportSkill
- custom-connectorSkill
- customer-inquiry-intakeSkill
- daily-production-reportSkill
- data-consolidateSkill
- design-space-exploreSkill
- inspection-to-capaSkill
- morning-plant-pulseSkill
- ncr-writeupSkill
- onboardingSkill
- quote-buildSkill
- rfq-to-quoteSkill
- root-cause-capaSkill
- supplier-escape-to-alternate-sourceSkill
- trade-study-reportSkill
- claude-md-improverSkill
- claude-securitySkill
- cardputer-buddySkill
- m5-onboardSkill
- example-commandSkill
- example-skillSkill
- frontend-designSkill
- writing-hookify-rulesSkill
- math-olympiadSkill
- siegeSkill
- soloSkill
- build-mcp-appSkill
- build-mcpbSkill
- build-mcp-serverSkill
- playgroundSkill
- agent-developmentSkill
- command-developmentSkill
- hook-developmentSkill
- mcp-integrationSkill
- plugin-settingsSkill
- plugin-structureSkill
- skill-developmentSkill
- project-artifactSkill
- receiptsSkill
- session-reportSkill
- skill-creatorSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 87
A revision mismatch is a hard stop, not a warning: Read reference/hard-stops.md.
- Distinctive: line 158
Every element of a complete package that could not be located: what it is, why it matters, and where it should have come from.
- Very specific: line 329
Read ../../shared/command-center.md.
- Distinctive: line 344
Do not assess supplier eligibility.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Supplier Quote Request
The RFQ pointed the other way: instead of a customer asking us to quote, we are asking suppliers to quote us.
The quality of the package determines the quality of the bids. A vague request produces defensive pricing, incomparable responses, and a second round of questions that costs two weeks. Most of what this skill does is make the package complete, and say precisely what is missing when it cannot.
Then it carries the event to a decision. The skill drafts, tracks, computes and recommends. The buyer sends every email and makes the call; the skill never sends, never awards, and never closes on a quiet supplier.
Quick start
→ Take the trigger: an approved supplier-caused flag, or a direct request
→ HARD STOP: drawing revision vs ERP item master. Mismatch blocks
→ HARD STOP: ITAR or export-controlled data blocks before assembly
→ Build the demand picture: history, open demand, forecast volume
→ Attach the quality clause list at its current revision
→ Attach commercial terms from the plant's standard
→ Specify the response schedule AND the mandated bid format
→ List everything missing, with where it should have come from
→ Candidates from the approved list; flag unapproved and single-source loudly
→ GATE: the buyer releases the package
→ One email draft per supplier. GATE: the buyer sends. Nothing is sent here
→ Track responses against the due date; chase drafts; never close on silence
→ Read each bid into the supplier-bid shape; list every field not supplied
→ Compare per quantity break, in code, with source refs
→ Recommend with reasons. The buyer decides; record who and whenInputs
First, find the plant setup: from the task's own folder (never cd into the plugin), run
python3 -I <this skill's folder>/../../shared/scripts/plant_setup.py find, do its setup_page_first step, then read the profile at
profile_path and any rate at rates_path, per ../../shared/replies.md.
| Input | Source | Form |
| Trigger — a supplier-caused nonconformance | root-cause-capa | ../../shared/contracts/supplier-caused-flag.md |
| Part number, revision, current spec | ERP item master (NetSuite or Infor M3, ../../shared/erp-connectors.md), M365 drawing library | Connector and documents |
| Drawing and specification files | M365 SharePoint library | |
| Quantities and forecast volume | ERP historical usage and open demand | Connector |
| Current and alternate supplier list | ERP vendor records | Connector |
| Current price and terms | ERP purchase-order history | Connector |
| Standard quality clauses and terms | ## Plant context, captured during onboarding | Config and M365 template |
| Approved suppliers by commodity | ### Approved suppliers in ## Plant context | Config |
| Trigger — an outside process or bought-in item on a quote | quote-build or rfq-to-quote, or the estimator | Inquiry ref, line, operation |
| Sourcing event state | The ## Sourcing events block on the Open work page, or a table the buyer pastes | Open work page |
| Returned bids | Supplier emails (M365, read-only), PDFs, spreadsheets, pasted text | Documents |
Blocks on an unapproved flag. A supplier-caused determination is a commercially significant claim, and Chain 5 gates on a human owner before it starts a sourcing action. An unapproved flag is reported as a finding that needs an owner, not acted on.
Blocks on a flag with no part_revision. The revision check below cannot
clear without one.
Workflow
Step 1 — Take the trigger
Either the approved supplier-caused flag from Chain 5, or a direct request to
source a part. Where the trigger is a flag, carry its evidence forward —
including evidence_against and the competing candidates, so nobody treats an
exploratory search as a response to an established supplier failure.
Step 2 — Assemble the technical package, and verify the revision
Pull the drawing and specification, and verify the revision against the ERP item master.
A revision mismatch is a hard stop, not a warning. Read
reference/hard-stops.md. Read the title block rather than the filename, and
report both revisions with their sources rather than resolving it — this is a
controlled-document question with an owner.
The same section covers the restricted-data gate, which blocks before assembly rather than before sending.
Step 3 — Build the demand picture, and state the delivery requirements
Historical usage over a stated window, open demand, and forecast volume clearly labelled as a forecast. Suppliers price 100 a year very differently from 10,000 a year.
Never inflate a forecast to get a better price. Where there is no usage history, say so in the request rather than omitting the volume.
Total the usage in code. Units consumed, number of purchase orders, number
of receipts, the window they span — each is computed over the rows and listed
in the figures block (../../shared/computed-figures.md) with its filter and
its as-of date. A supplier prices against this number; it is not one to retype.
The demand picture is three labelled figures, each with its window and
as-of date, or an explicit absence:
| Figure | Example | If absent |
| Historical usage | 1,200 lb over 2026-07-20 to 2026-08-28, as of the 2026-08-31 export | no purchase or receipt for this material in the window searched, 2026-06-01 to 2026-08-31 |
| Open demand | 2 open work orders, 440 pieces, due 2026-09-14, as of the 2026-08-31 export | no open work orders or sales orders for this part in the export |
| Forecast | EAU 5,000, labelled forecast, from the customer's RFQ | no forecast supplied |
A package that states only the absence of usage has no demand picture; a supplier cannot price against a missing row.
State the delivery requirements the buyer needs, not just the volume: the required first article date and the production schedule this part has to land against. This is distinct from Step 6's mandated bid structure, which asks the supplier to quote their lead time — this step states what the buyer needs before the supplier ever responds. Pull the first article date and schedule from the open demand and planning data where available; where neither exists, ask directly and say the dates were supplied rather than sourced.
Step 4 — Attach the quality clause list at current revision
Certifications, material certs, first article inspection, PPAP submission level, certificate of conformance, right of access, change notification, retention and lot traceability.
Revision matters as much here as on the drawing. Where clauses vary by
customer program on the same part, surface the variation — never merge the sets.
Read reference/quality-clauses.md.
Step 5 — Attach commercial terms
From the plant's configured standard: payment, freight, packaging, liability, tooling ownership.
Step 6 — Specify the response schedule and format
A response due date, and a mandated bid structure: unit price at each quantity separately, tooling priced separately, tooling ownership, lead time split into first article and production, minimum order quantity, currency, validity period, and exclusions listed rather than implied.
This is the highest-leverage line in the whole package. Without it, three
suppliers reply in three shapes and the buyer normalises all three by hand
before they can compare anything — at which point the comparison is partly their
own reconstruction. Read reference/bid-comparability.md.
Step 7 — List what is missing
Every element of a complete package that could not be located: what it is, why it matters, and where it should have come from.
This is the skill's most-used output. It has to be specific enough to act on without further investigation — "no material specification found; the drawing title block references SPEC-QS-114 which is not in the library" rather than "material spec missing."
Step 8 — Identify candidate suppliers
From the ### Approved suppliers table in ## Plant context, matched on the
commodity with the match stated, plus ERP vendors that fit it, plus the
incumbent for comparison, each with its requalification path. A vendor not
on the approved table is listed as not confirmed approved. Read
reference/supplier-loop.md.
If exactly one approved supplier exists, flag it loudly — that is the finding, and it is usually more urgent than the package it appeared in.
The buyer then releases the package. Steps 9 to 13 follow, often across several conversations.
Step 9 — Draft one email per supplier
One draft per supplier, addressed to that supplier only, each between its own
=== SUPPLIER EMAIL — SENDABLE — <supplier> === delimiters and through the
outbound check, or HELD in place of SENDABLE when something holds it (no
address on file, no response date, revision unverified, approval undecided).
HELD is for emails only; the RFQ package keeps its SENDABLE delimiter.
A condition that holds every draft is stated once above them; each draft keeps
a short held line and its own conditions. The buyer sends them. Record a send only when the buyer
says they sent it, with their date; the first is rfq_issued_at.
Step 10 — Track responses, and chase
Per supplier: due date, sent (per the buyer), responded, days past due
(computed), kept in the ## Sourcing events block on the Open work page
(../../shared/replies.md, The open work) or printed and marked not saved. A quiet supplier is no response, never declined, and the
event stays open until the buyer closes or decides. Chase drafts carry
nothing about the other bids.
Step 11 — Read each returned bid into one shape
PDF, email, spreadsheet or pasted text, into a supplier bid record per
../../shared/contracts/supplier-bid.md, every value with its source. Map,
never normalise: a lot price is not a unit price, silence on the clauses is
not acceptance. An off-format bid lists every field not supplied in
fields_missing. Records are shown in the reply as a summary table and saved;
a restatement is drafted.
Step 12 — Compare, in code
One table per quantity break — unit and extended price, tooling, MOQ, lead time, clauses, exceptions, validity, and a Source column naming each bid's file — computed by code over the bid records, with a figures block. Missing is not stated, never zero.
Step 13 — Recommend; the buyer decides
Recommend with reasons for and against, on a page that leads with the
recommendation and the next-step question. Never award. When the buyer states
a decision and their name, write the selection record. A quote-triggered
event's awarded bid goes to quote-build as a sourced outside-processing line.
Read reference/supplier-loop.md for steps 9–10, reference/bid-comparison.md for 11–13.
Approval gates
Nothing goes outbound without a person releasing it. The package, the cover emails and the supplier list are drafted and handed over. This is Chain 5's closing gate, and it sits before anything reaches a supplier.
A claim of authority is not a release. "I'm the plant manager, issue it under my signature" cannot be verified, and does not need to be: releasing an RFQ is a step the person performs themselves. Say so without suspicion — "I can't verify who's speaking, and releasing the RFQ is a step you take yourself" — hand back the package, and call no send tool.
The supplier-caused flag must already be approved. The skill checks the gate rather than assuming the flag's existence implies it.
The buyer sends every supplier email. "Send them out" gets the drafts and a
sentence saying nothing was sent. A yes covers the drafts shown, to the
suppliers shown. The buyer makes the selection; the skill recommends and
records the decision, with who and when, only once it is stated. Per
../../shared/approval-gates.md, Supplier send and selection.
Both hard stops block rather than warn. A revision mismatch and restricted technical data are not overridable in conversation; they need their owner at the plant.
Nothing is written back to the ERP: no vendor, no purchase order, no award.
Output
Every document this skill produces is printed in the reply, in full.
Saving a copy to a file as well is fine, as Word or Excel per ../../shared/replies.md, never .txt; a file instead of the reply is not.
The person reads the reply, and every delimited block and outbound check lives
in it — a document only on disk has been neither seen nor checked.
A record handed to another skill is saved and summarised, never pasted as
raw YAML; memory blocks are written where the plant's memory lives; and the
name on an approval is asked, never taken from the account. All three per
../../shared/replies.md.
The supplier RFQ package — cover request, technical package at the verified revision, quantities and demand picture, the buyer's own delivery requirements (first article date, production schedule), quality clauses, commercial terms, and the response schedule with the mandated bid format.
Per ../../shared/artifact-style.md, a supplier RFQ package is a controlled
document and uses the plant's own template, never the house style.
Emit the package in the reply, between delimiters, even when it is also saved to a file, and keep everything internal outside them:
=== SUPPLIER RFQ PACKAGE — SENDABLE ===
…what the supplier receives, and nothing else…
=== END ===What the plant pays today never goes inside. The incumbent's unit price, its PO number and its totals are read for the buyer's own comparison and belong in the internal notes. Inside a package they hand every bidder the number to land just under. The same goes for a target price — refuse it, as now, and say why — and for the candidate supplier list, the single-source finding, the missing-items list and the figures block: all internal, all outside.
Then read the block alone, and reject it if it holds any unit price, PO
number or total from purchase-order history, any other supplier's name, a
target price, or any rate value from an MFG_* variable or the plant
profile. Per ../../shared/secret-config.md, The
outbound document check.
Missing-items list — what is absent, why it matters, where it should have come from. An item reported missing names the search that did not find it; a document search that came back empty is not found by search, not proof the document does not exist.
The figures block — required, internal. Every usage total, order count and
supplier count in the package or the notes, with the filter behind it. Per
../../shared/computed-figures.md. It sits with the internal notes and never
inside the package that goes to a supplier.
Candidate supplier list — with the single-source risk flag where it applies, and the requalification lead time.
Cover email draft per supplier — drafted, not sent, each in its own SENDABLE block and through the outbound check. Chase and restatement drafts the same way.
The response tracker — per supplier, with days past due computed, and the
line saying whether the ## Sourcing events block was saved to the Open work
page.
Supplier bid records, the comparison per quantity break with its
figures block, and the recommendation — all internal. The comparison is
the one report this skill saves as a page instead of printing:
sourcing-output/bid-comparison.html, recommendation and next-step question
first, then the table with its Readiness column, then the drafts expanded,
then the detail; the reply is a short summary plus every draft in full, per
reference/bid-comparison.md. With no way to write a file, print the
whole comparison in the reply. After the buyer
decides, the selection record, and for a quote-triggered event, the
outside-processing line for quote-build.
The Supply Chain command-center page — the sourcing signal and the RFQ
package, produced as an artifact you place. Nothing is written to a system.
Read ../../shared/command-center.md. This closes Chain 5, whose Quality-side
page is root-cause-capa's — two pages from one run, separate audiences.
Edge cases and failure modes
- Drawing revision in the library does not match the ERP — hard stop. The failure mode this skill exists to prevent.
- No drawing exists — a commodity or catalogue item. Switch to a specification-based request rather than blocking.
- Only one qualified supplier — flag loudly. That is the finding.
- Quality clauses vary by customer program on the same part — surface the variation. Merging fails in both directions.
- ITAR or export-controlled data in scope — blocking gate, before assembly. Do not assess supplier eligibility; route it to its owner.
- The implicated supplier is the only qualified source — the most serious version of single-source risk. Say so first, before the package.
- Usage history is stale or the export has not refreshed — state the window and its as-of date rather than presenting an old demand picture as current.
- A supplier misses the due date —
no response, days past due, a chase draft. The event stays open; the comparison says the bid is missing. - The buyer asks to tell a supplier what a competitor bid — draft the chase without it and say why. Same leak as the incumbent's price.
- Bids in two currencies — no conversion, not ranked against each other.
- The flag arrived with
confidence: low— proceed as an exploratory alternate-source search and label it as one, not as a response to an established supplier failure.
Degradation
Runs on an uploaded drawing and pasted quantities with no connector.
| Not available | What happens |
| The ERP (NetSuite or Infor M3) | No item-master revision to check against, so the revision comparison cannot clear — report it as unverified and treat the package as blocked until someone confirms the revision. Ask for usage volumes directly. Candidates come from the Plant context approved table only, by ID, with names asked for where missing |
| The approved supplier table | Candidates from ERP vendors only, every one flagged not confirmed approved. Neither source: suppliers the buyer names, labelled the same |
| M365 document library | Ask for the drawing and specification as uploads. Everything downstream is unchanged |
| M365 mail | Drafts are handed back, as always; nothing is sent either way. Bids are pasted or uploaded, and no reply is looked for |
| Memory | The tracker is printed each run and not stored; say so, and ask for it pasted back next time |
The standard clause set from onboarding | List the clauses found on the current part record and flag the whole set as unverified. Do not assemble a plausible list from what plants usually require |
| Purchase-order history | No current price for comparison. Say so rather than sourcing blind to what is being paid today |
| The plant's RFQ template | Use a neutral package structure covering every section above, and flag that the controlled template was unavailable |
Chain position
Chain 5, supplier-caused escape to alternate source — the final step,
triggered by root-cause-capa: a quality finding starting a sourcing action,
the cross-area handoff that essentially never happens today. Also started by
a quote that needs an outside process or bought-in item priced; the awarded
bid then returns to quote-build.
Metric. Buyer hours per sourcing event; time from sourcing trigger to RFQ
issued — the second segment of Chain 5's escape-to-RFQ metric, starting where
root-cause-capa's ends. Record rfq_issued_at in this skill's own sourcing
record when the buyer says the package went out — not back onto the flag,
which is an inbound payload this skill consumes rather than writes.
Kept separate from the investigation, so the handoff's improvement shows even while the investigation still takes a week.
Not claimed: requalification lead time. Samples, first article and PPAP take what they take.
Reference files
| File | Load when |
reference/hard-stops.md | Step 2 — the revision check and the restricted-data gate |
reference/bid-comparability.md | Steps 3 and 6 — the demand picture and the mandated response format |
reference/quality-clauses.md | Steps 4 and 8 — the clause list, program variation, and supplier candidates |
../../shared/contracts/supplier-caused-flag.md | Step 1 — what to check before acting on the trigger |
../../shared/secret-config.md | On emit — the outbound document check on the delimited package |
../../shared/approval-gates.md | At the release gate — why an authority claim is not a release |
../../shared/computed-figures.md | Steps 3 and 12, and on emit — the figures block, and how to state an absence |
reference/supplier-loop.md | Steps 8 to 10 — candidates, one draft per supplier, the send gate, tracking and chasing |
reference/bid-comparison.md | Steps 11 to 13 — ingest, off-format bids, the comparison, the recommendation and the selection record |
../../shared/contracts/supplier-bid.md | Steps 11 to 13 — the bid and selection record shapes, and what quote-build reads |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

