agents-md-generator

• Commands are traced to their source — CI for ruff/pytest, Makefile for run/migrate — not inferred from the dependency list
• The test command is pytest -q as CI runs it, not a guessed pytest or make test that does not exist
• No command appears that cannot be traced to pyproject.toml, the Makefile, or the workflow
• Alembic migrations produce a convention or forbidden-action line (append-only / never edit a committed migration)
• A Do not section exists with concrete entries, not implied caution
• Output is at or under ~150 lines
• No virtue instructions appear — no 'clean', 'best practice', 'thorough', 'maintainable'
• Conventions included would each change agent behaviour if removed
• Audit mode is used — a verdict and diff, not a silent rewrite
• npm test is flagged as unverifiable: package.json has no test script
• The three virtue instructions are named for removal, with the reason that they change no behaviour
• What the file is missing is stated — real commands (build, lint), file placement, forbidden actions
• The line-count position is reported rather than the file being failed for length
• A concrete replacement or diff is offered, not just criticism
• Satellites are emitted as deltas only — each points at AGENTS.md for shared content
• No satellite repeats the command table, file placement, or conventions
• CLAUDE.md carries only Claude-Code-specific content; .github/copilot-instructions.md only Copilot-specific content
• If there is no genuinely tool-specific content, that is said rather than a file padded to justify itself
• No AGENTS.md is produced — the output is BLOCKED: need repo access
• No commands are invented from ecosystem convention
• The reason given is that every command must trace to a file in the tree
• Writes AGENTS.md at the repo root, under 60 lines, and modifies no other file
• Does not tell agents to run make test (there is no Makefile); the test command given is uv run pytest -m "not hardware" as in CI and CONTRIBUTING.md
• Gives ruff (uv run ruff format / uv run ruff check) as the formatter and linter, not black, resolving the README's stale instruction in favour of CI and CONTRIBUTING.md
• Tells agents not to run the tests marked hardware, because they need the live gateway on the plant network
• Forbids running scripts/deploy.sh and forbids editing, renaming or deleting anything in data/raw/
• Does not invent commands, tools or checks that appear nowhere in the repo (for example mypy, pre-commit, docker, or a Makefile target)
• Mentions that README.md is out of date (make test, black) as something for the humans to fix, without editing it
• Writes an AGENTS.md in the folder of roughly 10 lines (no more than about 15)
• Notices that make_report.R itself writes grades_final.csv, so the file tells the agent not to run make_report.R (or to ask first), not only 'don't edit grades_final.csv'
• Tells the agent not to change raw_marks.csv or grades_final.csv
• Gives the run order from notes.txt (Rscript clean_marks.R, then Rscript make_report.R) without adding commands or packages the folder does not mention
• Explains the change to the teacher in plain language, in a few lines
• Delivers the file without stopping to ask questions first

Should activate:

• Write an AGENTS.md for this repo
• Our AGENTS.md is stale — review it
• The agent keeps running commands that don't exist in our project
• Set this codebase up for agentic coding
• Create a CLAUDE.md for this project
• What should go in an agent instructions file?

Should leave to another skill:

• Write a CONTRIBUTING.md for this repo
• Document what this service does for new hires
• Reverse-engineer the architecture of this codebase
• Write a runbook for this service
• Review this design doc for gaps
• Design a supervised agent loop