wfukatsuwfukatsu/slide-forge/AGENTS.md

slide-forge

AGENTS.md5 starsChanged 56 days agoMITView on GitHub

Claude Code is the primary host and distribution path. Codex uses the same shared Python engine and the same skills/ through .agents/skills/ links. commands/, shared skills/, and references/workflow-contract.md own behavior; keep this file as a thin Codex compatibility layer. - Run Python commands from the repository root with .venv/bin/python. - Keep OAuth credentials, tokens, generated files, and caches out of Git. - Use GSLIDES_LANG=ja only when Japanese CLI output is useful; it does not affect generated content. - Before changing…

Reads credentials

Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository26 other files it gives its agents.
- forgeSkill
- analysis-template-creatorSkill
- b2b-account-mapsSkill
- calendar-slidesSkill
- current-state-analysisSkill
- drawio-diagramsSkill
- google-slidesSkill
- google-slides-templateSkill
- hearing-sheetSkill
- hearing-slidesSkill
- image-slotsSkill
- nexus-report-slidesSkill
- pptx-exportSkill
- pricing-strategySkill
- scalar-account-planning-sessionSkill
- scalar-account-planSkill
- scalar-ae-materialsSkill
- scalar-deal-intakeSkill
- scalar-nurture-intakeSkill
- scalar-product-slidesSkill
- scalar-proposal-slidesSkill
- settingsSkill
- slide-qaSkill
- slide-template-creatorSkill
- spreadsheetsSkill
- template-forgeSkill

The file

We split this AGENTS.md into the rules it gives an agent

9 rules. About 1 is common, 1 is distinctive, and 7 are very specific to this project.

≈1
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Very specific:
Run Python commands from the repository root with .venv/bin/python.
line 12
Common:
Keep OAuth credentials, tokens, generated files, and caches out of Git.
line 13
Very specific:
Use GSLIDES_LANG=ja only when Japanese CLI output is useful.
line 14
Distinctive:
Before changing an existing user-owned deck, follow the relevant skill's snapshot/version rule.
line 16
Very specific:
Use the matching skill and read its complete SKILL.md before acting:
line 21
Very specific:
After selecting a skill, do not preload the other skills or every reference it mentions.
line 52
Very specific:
Run the documented offline --dry-run/layout validation before API writes.
line 84
Very specific:
Never expose the contents of config/credentials.json, config/token.json, or API-key files.
line 85
Very specific:
Keep generated page fragments and QA thumbnails under ignored out/ paths.
line 91

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md 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.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

AGENTS.md

日本語

slide-forge agent instructions

Claude Code is the primary host and distribution path. Codex uses the same shared Python engine and the same skills/ through .agents/skills/ links. commands/, shared skills/, and references/workflow-contract.md own behavior; keep this file as a thin Codex compatibility layer.

Runtime

Run Python commands from the repository root with .venv/bin/python.
Keep OAuth credentials, tokens, generated files, and caches out of Git.
Use GSLIDES_LANG=ja only when Japanese CLI output is useful; it does not affect generated content.
Before changing an existing user-owned deck, follow the relevant skill's snapshot/version rule.

Skill routing

Use the matching skill and read its complete SKILL.md before acting:

Request	Skill
End-to-end deck workflow	`forge`
Registered template/master	`google-slides-template`
Deck without a corporate master	`google-slides`
Create/register a template	`template-forge`
Create/register a reusable single-slide content template	`slide-template-creator`
Current-state analysis / problem identification from supplied material	`current-state-analysis`
Value-based pricing: what to charge, a price increase or cut, price structure, discounts, a competitor's price cut — with the numbers as slides	`pricing-strategy`
Create/change an analysis-framework slide template	`analysis-template-creator`
Calendar slides from dated tasks/events (month calendar, day-by-day gantt, daily task list, weekly timetable, sprint calendar, year at a glance, deadline countdown, daily heatmap, shift roster)	`calendar-slides`
B2B stakeholder map / discovery mapping	`b2b-account-maps`
Per-customer activity plan / deal ledger (AE's action plan)	`scalar-account-plan`
Account Planning Session (annual/semi-annual stocktake and executive review)	`scalar-account-planning-session`
Materials for a single visit / internal approval materials (WPS, Deal Desk)	`scalar-ae-materials`
Organise minutes / email into per-stage records; work out what is still unconfirmed	`scalar-deal-intake`
Hearing sheet in Markdown / Excel / Google Spreadsheet, in either direction	`hearing-sheet`
Slides that collect the missing information (customer meeting / event)	`hearing-slides`
Design lead nurturing before a deal exists; segments, nurture tracks, content coverage	`scalar-nurture-intake`
Scalar product/company deck	`scalar-product-slides`
Scalar customer proposal	`scalar-proposal-slides`
Dense draw.io diagram	`drawio-diagrams`
Fill existing image frames	`image-slots`
Thumbnail visual QA	`slide-qa`
Export Google Slides to PPTX	`pptx-export`
Estimate/BOM spreadsheet; planning model whose numbers a deck links to	`spreadsheets`
Change the image-generation or output-destination switches	`settings`
Explain a nexus-architect project's reports / UI mocks as slides	`nexus-report-slides`

After selecting a skill, do not preload the other skills or every reference it mentions. Follow the progressive-loading table in references/workflow-contract.md.

Host-tool compatibility

Some shared skill documents retain Claude Code terminology. In Codex, apply these mappings:

Read for text: use rg, sed, or another read-only shell command.
Read for PNG/JPEG: use the local image-viewing tool and inspect the actual pixels; filenames and API success are not visual QA.
Write / Edit: use apply_patch for repository files.
Bash: use the shell execution tool with the repository root as cwd.
Grep / Glob: use rg / rg --files first.
WebFetch / WebSearch: use the available web tool. Prefer official, primary sources for product facts and technical documentation.
AskUserQuestion / ask_question: ask in chat. Present mutually exclusive choices as a numbered list and wait for the user's number; accept comma-separated numbers for multi-select. Do not silently choose when the skill requires an approval gate.
Task / Subagent / Parallel: use parallel agents only when the current host and session instructions explicitly permit them. Otherwise follow the sequential fallback in references/parallel-generation.md.
${CLAUDE_PLUGIN_ROOT}: resolve to this repository root. Never depend on the variable being set under Codex.

Detailed Codex setup and known differences are in references/codex-compatibility.md.

Safety and validation

Run the documented offline --dry-run/layout validation before API writes.
Never expose the contents of config/credentials.json, config/token.json, or API-key files.
accounts/ holds per-customer sales ledgers — named individuals and judgements about them. It is ignored by Git; never commit it, never paste it into a customer-facing artifact, and never share the 00_活動計画 / 90_社内 Drive folders with a customer or partner.
Keep generated page fragments and QA thumbnails under ignored out/ paths.
A successful API response is not visual QA. When QA is selected, inspect the thumbnails and run scripts/cleanup_qa.py before reporting completion.

Download now (ZIP)

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

What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free