pack-new
Skill0 starsChanged 25 days agoMITView on GitHub
Create a new Pack — guided flow through SPF: choose domain, name Pack,
1 other project uses a version of this.
Adopted by
Most recently taken up by
- TserenTserenov/FMT-exocortex-templatetheir own version, first seen Nov 2, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- TserenTserenov/FMT-exocortex-template★ 64same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository51 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- agent-faultSkill
- apply-capturesSkill
- archgateSkill
- artifactorSkill
- audit-docsSkill
- audit-installationSkill
- author-modeSkill
- bottleneck-pickSkill
- check-secretSkill
- consentSkill
- day-closeSkill
- day-openSkill
- decomposeSkill
- diagnoseSkill
- discovery-sessionSkill
- extendSkill
- fpfSkill
- integration-gateSkill
- iwe-bug-reportSkill
- iwe-platform-redteamSkill
- iwe-restoreSkill
- iwe-rules-reviewSkill
- iwe-updateSkill
- keSkill
- kimi-peer-writerSkill
- local-llmSkill
- month-closeSkill
- org-devSkill
- pack-creatorSkill
- peer-conversationSkill
- personal-guide-renderSkill
- personal-guide-startSkill
- platform-bottleneckSkill
- repo-newSkill
- restore-exocortexSkill
- run-protocolSkill
- setup-wakatimeSkill
- skill-creatorSkill
- strategy-sessionSkill
- skill-idSkill
- thinkSkill
- transcribeSkill
- vdvSkill
- verify-hypothesesSkill
- verifySkill
- week-close-pilotSkill
- week-closeSkill
- wp-newSkill
- w-reflectionSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
16 rules. About 0 are common, 5 are distinctive, and 11 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
11
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 27
Does not fill Pack content beyond the starter SoTA Sheet from Step 1.5 (see Step 4) — everything else requires /ke and manual work per SPF …
- Very specific: line 34
Check that SPF/ and FPF/ exist in the IWE working directory.
- Distinctive: line 36
If absent — notify the user and propose the commands:
- Very specific: line 46
If the repositories exist — record the path to SPF/pack-template/ for Step 4.
- Distinctive: line 68
If the user is unsure — help find practitioners and Artifacts through follow-up questions (up to 2 additional questions).
- Distinctive: line 76
Before choosing a name (Step 2), quickly verify: how practitioners themselves name and describe this Practice — not from the agent's …
- Distinctive: line 78
Ask the user (or find independently if the user is not an expert):
- Very specific: line 85
If the user says "I do not have a source at hand" — do not block, continue.
- Distinctive: line 87
Sync point: after Step 3 (Bounded Context) — check whether the domain boundary has shifted enough that the source and claims need to be …
- Very specific: line 140
If pack_id_slug changed: verify that PACK-{new_slug}/ does not already exist (git mv into an existing directory moves files INSIDE it …
- Very specific: line 141
If pack_id_code changed or is being set for the first time: check for collision with another Pack's code in IWE (see code anti-example in …
- Very specific: line 143
Verify the replacement: grep -c '{{PACK_ID}}' <file> after replacement must be 0, and the number of ### {new_code}.D.NNN headings in the …
- Very specific: line 146
Commit the changed files (by explicit path list, not git add -A) with a message of the form feat: finalize Pack name → PACK-{new_slug} …
- Very specific: line 174
Create ${IWE_WORKSPACE:-$HOME/IWE}/PACK-{slug}/ with the following structure:
- Very specific: line 268
All 5 points must be filled substantively (not with placeholders) before removing the Maturity: seed line.
- Very specific: line 332
If no source was provided (Step 1.5, user skipped) — do not create the file.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Pack New — Creating a New Pack
Creating a Pack for the domain: $ARGUMENTS
What this Skill does
- Checks for the Base repositories (FPF, SPF) and clones them if absent
- Guides through SPF §01 (domain selection)
- Checks domain/name recognizability against one external source (SoTA-Sheet-lite)
- Proposes a provisional Pack name per SPF criteria, records rejected candidates (PFAD-lite), and finalizes the name after the first Distinctions
- Refines the Bounded Context (SPF §02)
- Creates the directory structure and starter files from SPF/pack-template
- Shows the content Roadmap with time estimates
What this Skill does NOT do
- Does not fill Pack content beyond the starter SoTA Sheet from Step 1.5 (see Step 4) — everything else requires
/keand manual work per SPF §03–11 - Proposes creating a GitHub repository by command only — does not create it automatically
Step 0. Check Base repositories (FPF + SPF)
Check that SPF/ and FPF/ exist in the IWE working directory.
If absent — notify the user and propose the commands:
(cd ${IWE_WORKSPACE:-$HOME/IWE}
cp -R "${IWE_TEMPLATE:-$HOME/IWE/IWE}/seed/SPF" SPF # if SPF/ is missing
git -C SPF init -q
gh repo clone ailev/FPF FPF -- --depth=1 # if FPF/ is missing
)If the repositories exist — record the path to SPF/pack-template/ for Step 4.
Also record the Work Product in whose context /pack-new was launched (from the active Session — the WP Gate must already be passed before calling this Skill per Pack Creation Gate). If the Pack is created outside a Work Product context — record wp: —. This value is needed for the wp: field in the .pfad-decision.md frontmatter in Step 4.
Step 1. Domain ≠ Topic (SPF §01)
Ask the user no more than 3 questions (all at once, in a single message):
- Who practices this domain? (specialty, profession, specific Role)
- What do they produce? (Artifacts, Work Products — specific documents, systems, decisions)
- How do they typically fail? (3–5 failure modes — what goes wrong in this Practice)
If the answer is "I am interested in topic X", explain the Distinction:
Topic = area of interest (no own methods or Artifacts). Domain = Practice with methods, Work Products, and failure modes.
Test: "Are there people who DO this professionally? What do they produce?"
Examples: "machine learning" — topic. "ML systems development" — domain (has: ML engineers, models, quality metrics, failure modes). "Systems Thinking" — topic. "Systems analysis" — domain.
If the user is unsure — help find practitioners and Artifacts through follow-up questions (up to 2 additional questions).
Step 1.5. Domain sources (SoTA-Sheet-lite)
Source of principle: FPF
E.4.DPF(source pack — step 2 of 11, before drafting patterns) +G.2lightweight variant ("1-page SoTA Sheet", informative). The fullG.2(CorpusLedger/ClaimSheets/BridgeMatrix) is excessive for a personal Pack.
Before choosing a name (Step 2), quickly verify: how practitioners themselves name and describe this Practice — not from the agent's memory, but from one external source.
Ask the user (or find independently if the user is not an expert):
- One authoritative source for this Practice — a book, method, school, or standard (not a blog)
- 2–4 claims from that source worth carrying into the Pack
- What supports them — a quote, page number, or section
This is not a full research pass — the goal is narrow: verify that the domain and the future name (Step 2) are grounded in real Practice. Optional, if time permits: applicability boundaries of the source (validity region), what is explicitly rejected, when the source will become outdated (freshness).
If the user says "I do not have a source at hand" — do not block, continue. Record this explicitly in Step 4 (sota_sources: none in the Manifest), do not stay silent.
Sync point: after Step 3 (Bounded Context) — check whether the domain boundary has shifted enough that the source and claims need to be supplemented.
Step 2. Provisional Pack name (SPF §01 §4)
Pack name = a noun recognizable to practitioners of the domain.
Criteria (all required):
- Specific: excludes neighboring domains (not "management", but "product management")
- Broad: covers the core methods, not just one tool
- Recognizable: a practitioner immediately understands what it refers to
- Slug: Latin characters, kebab-case, ≤30 characters
Propose 2–3 options with explanations, then let the user choose.
Format: PACK-{pack_id_slug} (examples: PACK-product-management, PACK-system-analysis, PACK-digital-marketing).
Reference examples from IWE: PACK-digital-platform, PACK-education, PACK-personal, PACK-verification.
Anti-examples:
PACK-everything— too broadPACK-jira— a tool, not a domainPACK-notes— no Practice or Artifacts
Short code (pack_id_code, WP-474 phase 3 fix). Separate from pack_id_slug (kebab-case, used as the directory name) — a short mnemonic code of 2–4 uppercase Latin letters used as a prefix in entity codes ({{PACK_ID}}.D.NNN → e.g. DP.D.NNN). These are two different identifiers — do not confuse them: existing Packs use the short code in this role (PACK-digital-platform → code DP), not the full slug.
Code proposal algorithm: take the significant words of the slug (excluding prepositions/conjunctions) → if the domain is a single word, take the first 2–4 letters (education → EDU); if it consists of multiple words, take the first letter of each significant word in uppercase (product-management → PM; digital-platform → DP). Propose the resulting code plus one alternative (e.g., first letters of the first word plus the first letter of the second word, if the initials yield fewer than 2 significant words) — let the pilot choose or adjust. If the proposed code is already taken (see anti-example below) — do not blindly ask the pilot again; instead, immediately propose the next option produced by the algorithm (e.g., add a third letter), and only ask the pilot directly if all options are exhausted. The code is also provisional — it is finalized together with the slug (see "Name finalization" below).
Code anti-examples:
- Collision with an existing Pack code in IWE — check BEFORE recording:
find ${IWE_WORKSPACE:-$HOME/IWE} -maxdepth 4 -path "*/PACK-*/00-pack-manifest.md" -o -path "*/PACK-*/pack/*/00-pack-manifest.md" | xargs grep -l "^pack_id: <code>$" 2>/dev/null(Packs can be flat —PACK-X/00-pack-manifest.md— or nested —PACK-X/pack/X/00-pack-manifest.md; onefindbranch does not cover both variants) - Code longer than 4 characters — loses compactness in compound IDs (
DIGPLAT.D.001instead ofDP.D.001)
Provisionality (PFAD-lite, source — FPF E.4.PFAD/F.18). The name chosen here (slug + code) is not final. The directory PACK-{pack_id_slug}/ is created immediately under this slug (Step 4), but:
00-pack-manifest.mdgetsname_status: provisional;- rejected domain/name/boundary candidates are recorded in
.pfad-decision.md(Step 4) with the reason for rejection — a decision record analogous to an ADR, living inside the Pack and traveling with it on moves or renames; - Distinctions added to
01B-distinctions.mdbefore finalization (roadmap phase 1, Step 5) receive the heading### {{PACK_ID}}.D.NNN: <Name>—{{PACK_ID}}is a placeholder, replaced at finalization with the short code (pack_id_code), NOT withpack_id_slug; - name finalization is a separate step after content accumulates (see below), not an automatic transition.
While discussion is ongoing (this step and, if needed, Step 3) — maintain a draft list of rejected domain/name/boundary candidates with reasons for rejection (in the working dialog context). This draft is transferred to the tables in .pfad-decision.md in Step 4 — if the Session is interrupted between Step 2 and Step 4, the rejected candidates must not be lost.
Name finalization
Triggered by one of two conditions (record which one in .pfad-decision.md field finalization_trigger):
pilot-requested— the user explicitly requests the name to be fixed, at any time.agent-proposed-after-N-distinctions— the agent proposes finalization after at least 3 Distinctions appear in01B-distinctions.md(not earlier — one or two are insufficient to judge whether the name fits the content).
The two independent identifiers are finalized together but renamed differently: pack_id_slug (kebab-case) — rename the DIRECTORY via git mv; pack_id_code (short mnemonic code) — replace the PLACEHOLDER {{PACK_ID}} in Distinction content. Do not confuse: {new_slug} in the steps below is NOT the same as what replaces {{PACK_ID}}.
Paths in provisional_distinction_files are stored relative to the Pack root (e.g. 01-domain-contract/01B-distinctions.md, without the PACK-{slug}/ prefix) — always resolve them to a real file path using the CURRENT directory name of the Pack at the time of the operation. Therefore, after git mv (step 2 below) the paths from the list are already correct without separate modification — look them up as PACK-{new_slug}/<path-from-list>.
Procedure:
- The user confirms the current
pack_candidate/pack_id_slug/pack_id_codeor provides new ones (independently — the slug may stay the same while the code changes, and vice versa). - If
pack_id_slugchanged: verify thatPACK-{new_slug}/does not already exist (git mvinto an existing directory moves files INSIDE it instead of renaming) — on collision, stop and ask the pilot for a different slug. Otherwise rungit mv PACK-{old_slug} PACK-{new_slug}. - If
pack_id_codechanged or is being set for the first time: check for collision with another Pack's code in IWE (see code anti-example in Step 2). On collision — stop and ask for a different code. - For each path in
provisional_distinction_files, resolving it asPACK-{new_slug}/<path>(see above): count occurrences of{{PACK_ID}}BEFORE replacement (grep -c '{{PACK_ID}}' <file>= N), then replace all occurrences of{{PACK_ID}}→{new_code}(the short code, NOT the slug) in Distinction headings. - Verify the replacement:
grep -c '{{PACK_ID}}' <file>after replacement must be0, and the number of### {new_code}.D.NNNheadings in the file must equal N. Any discrepancy (remaining{{PACK_ID}}or heading count ≠ N) — do not commit; show the diff to the user. - Update all locations where names are materialized:
pack_id(={new_code}) andpack_namein00-pack-manifest.md; ifpack_id_slugchanged — the# PACK-{new_slug}heading in the Pack'sCLAUDE.md, thepack_id_slug/pack_candidate/pack_id_codefields in.pfad-decision.md, rename06-sota/{old_slug}-sota-sheet.md→06-sota/{new_slug}-sota-sheet.md(if the file exists), rename.iwe-runtime/state/spf/{old_slug}.yaml→.iwe-runtime/state/spf/{new_slug}.yaml(if it exists —/pack-creatorhas already been run, WP-474 phase 3; otherwise it will not find the progress on the next run and will silently restart from Step 0). If a GitHub repository was created under the old slug in Step 4 — notify the pilot that it must be renamed separately (gh repo rename); this Skill does not do that automatically. - Clear
provisional_distinction_files, setname_status: finalizedin the Manifest andstatus: finalizedin the.pfad-decision.mdfrontmatter, fill in the "Final choice" section of.pfad-decision.md(decided_by,proposed_by; if the**Kind:**line was left blank from Step 3 — ask about Kind now and fill it in). Lexicon backstop: ifontology.md§2 Domain Glossary remains placeholder-only (terms from Step 3 were not transferred) — fill it in now. - Commit the changed files (by explicit path list, not
git add -A) with a message of the formfeat: finalize Pack name → PACK-{new_slug} (code {new_code}).
Step 3. Bounded Context (SPF §02)
Fill in three fields together with the user:
| Field | Content |
| What is included | 3–5 key methods and Practices of the domain |
| What is excluded | Neighboring domains (boundary) |
| Key terms | 5–7 domain-specific terms (Ubiquitous Language) |
Result → recorded in 01-domain-contract/01A-bounded-context.md
Term materialization (WP-474 phase 5, flag O of coordinate D6). Key terms from the table above are recorded NOT only in 01A: each term becomes a row in ontology.md §2 Domain Glossary (Term RU/EN, Definition in 1–2 sentences, Parent Concept from the SPF base ontology — see SPF/pack-template/ontology.md §1). Without this, /verify pack will correctly report the lexicon as unfilled — 01A verify does not read it.
Kind of the main concept (WP-474 phase 5, flag S of coordinate D6). At this point, ask: "What is the base Kind of the main concept of the domain — U.Method (way of acting), U.System (carrier), U.Episteme (Knowledge), or another from the SPF base ontology?" Discussed and rejected candidates go into the ## Kind table in .pfad-decision.md (Step 4); the accepted decision goes into the **Kind:** line in "Final choice" immediately — it does not wait for name finalization.
Source sync point (Step 1.5): if the domain boundary has changed noticeably here — go back and supplement the SoTA Sheet before Step 4.
Step 4. Scaffold the structure
{slug} throughout = the {pack_id_slug} chosen in Step 2.
Create ${IWE_WORKSPACE:-$HOME/IWE}/PACK-{slug}/ with the following structure:
PACK-{slug}/
├── README.md ← name + one sentence about the domain
├── REPO-TYPE.md ← type: Pack, upstream: FPF + SPF
├── CLAUDE.md ← agent instructions for this Pack
├── .pfad-decision.md ← PFAD-lite: rejected domain/name/boundary candidates (Step 2)
├── 00-pack-manifest.md ← metadata + entity index
├── ontology.md ← domain terms (Ubiquitous Language)
├── 01-domain-contract/
│ ├── 01A-bounded-context.md ← from Step 3
│ └── 01B-distinctions.md ← key Distinctions (starter template)
├── 02-domain-entities/ ← entities: roles, methods, Work Products
├── 03-methods/ ← Practice methods
├── 04-work-products/ ← Work Products
├── 05-failure-modes/ ← typical errors
├── 06-sota/
│ └── {slug}-sota-sheet.md ← from Step 1.5 (if a source was provided)
└── 07-map/ ← domain mapFill in the starter files:
README.md — one line describing the domain.
REPO-TYPE.md:
# Repository type
**Type**: `Pack`
**Source-of-truth**: yes
## Scope
{domain name and what it covers}
## Upstream dependencies
- [Bundled SPF](https://github.com/iwesys/IWE/tree/main/seed/SPF) — Second Principles Framework
- [ailev/FPF](https://github.com/ailev/FPF) — First Principles Framework
## Non-goals
- Does NOT contain code and Configurations (→ DS)
- Does NOT contain plans and registries (→ DS/governance)CLAUDE.md — minimal, contains:
# PACK-{slug}
Source-of-truth for the domain: {name}.
Structure: SPF/pack-template. Upstream: FPF, SPF.
When working with this Pack: read 00-pack-manifest.md for navigation.
While `name_status: provisional` in the Manifest: each new Distinction in `01-domain-contract/01B-distinctions.md` receives the heading `### {{PACK_ID}}.D.NNN: <Name>` (the placeholder `{{PACK_ID}}`, not the real Pack code), and the path `01-domain-contract/01B-distinctions.md` is added to `provisional_distinction_files` in `.pfad-decision.md` (if not already present — do not duplicate). If `01B-distinctions.md` contains 3 or more Distinctions — propose name finalization to the user (see `pack-new/SKILL.md` Step 2 "Name finalization" in IWE).01-domain-contract/01A-bounded-context.md — from Step 3.
01-domain-contract/01B-distinctions.md — template (heading per Distinction, aligned with the actual format used in active Packs, e.g. PACK-digital-platform — the table from the previous version of this template matched neither the provisional-placeholder heading format used below nor the actual format of active Packs; corrected in WP-474 phase 3. SPF/pack-template/01-domain-contract/01B-distinctions.md uses a different notation — ### [D.001] Name without a Pack code prefix — the upstream template diverges from Practice; noted separately, not resolved by this fix):
# Key Distinctions of {domain}
> Source: FPF A.7 (Strict Distinction)
> Criterion: if two terms are frequently confused — this is a Distinction.
### {{PACK_ID}}.D.001: <Name>
**Definition A:** ...
**Definition B:** ...
**Test:** ...While name_status: provisional (see Step 2) — each added Distinction uses the heading ### {{PACK_ID}}.D.NNN: <Name> (not the real Pack code), and the path 01-domain-contract/01B-distinctions.md is added to provisional_distinction_files in .pfad-decision.md (if the path is not already there — do not duplicate on repeated Distinctions in the same file). After name finalization (Step 2 "Finalization") {{PACK_ID}} is replaced with the short code (pack_id_code) — NOT the slug — in one verifiable step.
Seed/mature (WP-474 phase 3, peer session 2026-07-09-14-seed-mature-spf-state). Every new Distinction is a draft (seed) by default. Mark it with a line immediately below the heading, not with a suffix in the heading itself (a suffix in the heading breaks the stability of markdown anchors during the seed→mature transition — links from other files to #packid-d-001-name-seed will break when the label is removed). The field is named **Maturity:**, not **Status:** — in existing Pack cards **Status:** is already used for SoTA status (current/deprecated/hypothesis, a different axis), and reusing the field name would create confusion for any future tool that searches for Status across a Pack:
### {{PACK_ID}}.D.001: <Name>
**Maturity:** seed
**Definition A:** ...mature = absence of the **Maturity:** line (symmetric with the rest of the format — the "normal" state is not marked explicitly).
seed → mature transition — not automatic; triggered by the author's request or by an agent proposal when a Distinction appears stable. "mature-lite" checklist (lightweight version of FPF E.8 — the 13 canonical pattern sections are excessive for a short Pack Distinction, following the same lightweight principle as in phases 1/2):
- Problem/motivation — why this Distinction is needed, what gets confused without it
- Forces — what alternative or tension it resolves (without this point, maturity is a declaration: it is unclear whether the author worked through the conflict or the Distinction does not exist in practice)
- Example — at least one real worked example, not an abstraction
- Common mistake — at least one real misuse case, not a placeholder
- Consequences — what breaks in Practice if the Distinction is ignored
All 5 points must be filled substantively (not with placeholders) before removing the **Maturity:** seed line. If a point does not apply — state explicitly why; do not leave it blank.
.pfad-decision.md — PFAD-lite decision record (SPF E.4.PFAD, lightweight version — 3 tables instead of the full format), created together with the scaffold in Step 4, populated from the results of Step 2 (and Step 3, if the boundary was discussed):
---
wp: "{Work Product in whose context the Pack is created}"
pack_candidate: "{name proposed in Step 2}"
pack_id_slug: "{slug}"
pack_id_code: "{short code, 2-4 letters}"
created: "{YYYY-MM-DD}"
status: provisional
finalization_trigger: ""
provisional_distinction_files: []
---
# PFAD-lite: {pack_candidate}
## Domain
| Candidate | Why rejected |
|-----------|-------------|
## Name
| Candidate | Why rejected |
|-----------|-------------|
## Boundary
| Candidate | Why rejected |
|-----------|-------------|
## Kind
| Kind candidate | Why rejected |
|----------------|-------------|
## Final choice
**Domain:** ...
**Name (slug):** ... (was provisional: "...")
**Code:** ... (was provisional: "...")
**Boundary:** ...
**Kind:** ...
**decided_by:** pilot
**proposed_by:** agent | pilotFill in the tables with candidates that were actually discussed in Step 2 (and Step 3) — do not invent rejected candidates after the fact; if the user immediately agreed with the first proposal, the tables remain empty. The "Final choice" section is filled in at finalization (Step 2 "Name finalization"), not in Step 4.
The ## Kind table and the **Kind:** line (WP-474 phase 5, D6-settlement). Kind is the base type of the domain's main concept per the SPF base ontology (U.Method / U.System / U.Episteme / ... — see SPF/pack-template/ontology.md §1). The Kind decision is made in Step 3 (the question is asked there). The ## Kind table is a registry of rejected Kind alternatives (it may remain empty per the general rule of not inventing candidates after the fact); the table itself is NOT evidence of a decision. The evidence of settlement is only the filled **Kind:** line in "Final choice" (checked by /verify pack, coordinate D6). Exception to the rule in the paragraph above: unlike Domain/Name/Boundary, the **Kind:** line is filled immediately upon the decision (Step 3) and does not wait for name finalization.
06-sota/{slug}-sota-sheet.md — from Step 1.5:
# SoTA Sheet: {source}
**Claims:** {2-4 claims carried into the Pack}
distinction: {X} vs {Y}
**Evidence:** {quote/page/section}
**Validity region:** {where it applies, where it does not — optional}
**Rejected:** {what is explicitly rejected and why — optional}
**Freshness:** {when to revisit — optional}distinction: X vs Y lines (0..N, optional) — Distinction candidates that the source draws explicitly. The format is strict (the line begins with distinction:, separator vs), position is free — the pack-creator parser looks for such lines anywhere in the sota-sheet file. The /pack-creator assembly mode uses them to deterministically build a candidate checklist (WP-474 phase 6). They are not required in Step 1.5 — they can be added later, including via LLM-fallback pack-creator with author confirmation.
If no source was provided (Step 1.5, user skipped) — do not create the file; record this only in 00-pack-manifest.md (sota_sources: none).
ontology.md — take the template from SPF/pack-template/ontology.md, then IMMEDIATELY fill in §2 Domain Glossary with the terms established in Step 3 ("Term materialization"): row by row — Term (RU/EN) / Definition / Parent Concept (SPF) from the discussion. Do not leave template placeholders _Term 1_ / _TBD_ — /verify pack (flag O of coordinate D6) treats placeholder lines as an unfilled lexicon.
00-pack-manifest.md — take the template from SPF/pack-template/00-pack-manifest.md, fill in pack_id = the pack_id_code from Step 2 (the short mnemonic code, e.g. DP, NOT pack_id_slug and NOT PACK-{slug} — the discrepancy was found and corrected in WP-474 phase 3: real Packs use the short code in this field), pack_name = pack_candidate. In the ## Metadata block, immediately after pack_name, add two new lines (they are not yet in the upstream template — do not confuse this with filling an existing placeholder; adding them to the template, if needed for all Packs, is a separate SPF decision, not pack-new's responsibility):
sota_sources: none | groundedfrom the result of Step 1.5 —noneif no source was provided,groundedif a source and claims were gathered.name_status: provisional | finalizedfrom the result of Step 2 — alwaysprovisionalat scaffold time (Step 4); transitions tofinalizedonly at name finalization (Step 2 "Finalization").
Then initialize the repository and install the CI guard:
(cd ${IWE_WORKSPACE:-$HOME/IWE}/PACK-{slug}
git init
# Install CI guard (ID collision detector)
IWE_TEMPLATE="${IWE_TEMPLATE:-$HOME/IWE/IWE}"
if [ -d "$IWE_TEMPLATE/pack-templates/.github" ]; then
cp -r "$IWE_TEMPLATE/pack-templates/.github" .
fi
git add -A
git commit -m "feat: initial scaffold PACK-{slug} (SPF/pack-template) + CI guard R4"
)The CI guard is a GitHub Action that checks ID uniqueness (DP.M.NNN, AR.NNN, etc.) on every push/PR and blocks merging on collision.
Optionally — create on GitHub:
gh repo create {GITHUB_USER}/PACK-{slug} --private --source="${IWE_WORKSPACE:-$HOME/IWE}/PACK-{slug}" --pushStep 5. Content Roadmap
Show the user the plan — what to do next:
═══════════════════════════════════════════════════════
PACK-{slug} — content roadmap
═══════════════════════════════════════════════════════
Current state: scaffold complete. Pack is empty — value appears after phase 1.
[ ] Phase 1. DISTINCTIONS (SPF §03) ← start here
File: 01-domain-contract/01B-distinctions.md
Goal: 7-10 key domain Distinctions
Time: ~1-2h
How: list what is frequently confused; check each against FPF A.7
Tool: /ke — capture Distinctions while working with the domain
[ ] Phase 2. ENTITIES (SPF §04)
File: 02-domain-entities/
Goal: list Roles, Work Products, methods — without detailed descriptions
Time: ~1-2h
How: answer the question "who acts, what do they produce, how do they verify?"
[ ] Phase 3. METHODS (SPF §07)
File: 03-methods/
Goal: describe the key Practice methods
Time: ~2-4h
Template: for each method — inputs, outputs, quality criteria
[ ] Phase 4. WORK PRODUCTS (SPF §07)
File: 04-work-products/
Goal: describe the Practice Artifacts
Time: ~1-2h
Template: name, description, Definition of Done
[ ] Phase 5. FAILURE MODES (SPF §08) ← high value
File: 05-failure-modes/
Goal: 5-10 typical domain errors
Time: ~1h
Format: FM.NNN — cause, signal, how to avoid
[ ] Phase 6. SoTA — expand sources (SPF §09)
File: 06-sota/
Goal: gather the first source (if Step 1.5 was skipped, sota_sources: none)
or deepen sources beyond the starter SoTA Sheet — as the Pack grows
Time: ~1-2h
═══════════════════════════════════════════════════════
Tools for content work:
/ke — capture Knowledge into the Pack during work
/fpf — verify entity correctness per FPF A.*
/verify pack — baseline adequacy Assessment across 11 coordinates
E.4.DPF.DA (expect `seedOnly` for a fresh scaffold:
some coordinates will honestly be below threshold — normal for a template).
Required run — in /pack-creator Step 4.
SPF/process/03-distinctions-work.md — detailed process for phase 1
SPF/process/08-failure-modes-extraction.md — process for phase 5
═══════════════════════════════════════════════════════
Principle: a Pack is not filled in a single sitting.
First value — 7-10 Distinctions (phase 1, 1-2h).
The Pack then grows through work with the domain via /ke.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.

