AI-PM-essential-skills
AGENTS.md6 starsChanged 5 months agoMITView on GitHub
These rules apply throughout this repository. More specific AGENTS.md files may add stricter local rules but must not weaken these requirements. ContextPort lives in context-port/. Do not create a separate repository for it. Do not modify existing AI-PM skills while working on ContextPort without explicit human approval. Codex may work autonomously on: Stop for human approval before: When a decision is required, return exactly this structure: Every meaningful phase requires one saved prompt, one feature branch, logical commits, tests or…
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 33,315 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 repository5 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- prd-firstSkill
- concise-rewriterSkill
- context-auditorSkill
- token-cost-estimatorSkill
The file
We split this AGENTS.md into the rules it gives an agent
11 rules. About 1 is common, 10 are distinctive, and 0 are very specific to this project.
≈1
Common · estimate
Many files like this have a rule of this kind.
≈10
Distinctive · estimate
This author's own choices. Worth a look.
0
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
- Distinctive: line 26
Stop for human approval before:
- Distinctive: line 41
When a decision is required, return exactly this structure:
- Distinctive: line 55
Never invent APIs, export fields, browser selectors, or results.
- Distinctive: line 57
Never report success based only on an attempted action.
- Distinctive: line 58
Never silently discard or truncate content.
- Common: line 59
Never commit personal conversations, exports, passwords, cookies, or sessions.
- Distinctive: line 60
Use synthetic data in committed tests.
- Distinctive: line 65
After two failed attempts, return an honest failure report.
- Distinctive: line 69
Every meaningful phase requires one saved prompt, one feature branch, logical commits, tests or evaluation evidence, one draft pull …
- Distinctive: line 71
Use logical commits appropriate to the phase, such as:
- Distinctive: line 79
Never push directly to main.
≈ “Common” is an estimate: at least 1 in 5 of the 42,283 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.
Files
Codex operating rules
These rules apply throughout this repository. More specific AGENTS.md files may add stricter local rules but must not weaken these requirements.
ContextPort scope
ContextPort lives in context-port/. Do not create a separate repository for it. Do not modify existing AI-PM skills while working on ContextPort without explicit human approval.
Autonomous work
Codex may work autonomously on:
- Reversible code and documentation changes
- Architecture
- Tests
- Synthetic fixtures
- Refactoring
- Linting
- Type checking
- Git commits
- Pushing the current feature branch
- Creating a draft pull request
Human approval gates
Stop for human approval before:
- Accessing a real Claude or ChatGPT export
- Launching browser automation
- Writing anything into Claude or ChatGPT
- Resolving ambiguous project mappings
- Truncating or summarising conversations
- Adding production dependencies
- Sending private data over the network
- Deleting or overwriting content
- Modifying existing AI-PM skills
- Merging a pull request
- Publishing a release
- Making a material product decision with multiple valid options
When a decision is required, return exactly this structure:
DECISION REQUIRED
Question:
Evidence:
Option A:
Option B:
Recommendation:
Reversibility:Reliability rules
- Never invent APIs, export fields, browser selectors, or results.
- Label assumptions
VERIFIED,INFERRED,UNKNOWN, orUNSUPPORTED. - Never report success based only on an attempted action. Require observable evidence.
- Never silently discard or truncate content.
- Never commit personal conversations, exports, passwords, cookies, or sessions.
- Use synthetic data in committed tests.
- Preserve original project and conversation titles.
- Make reruns idempotent.
- Report everything not copied.
- Allow a maximum of two repair attempts after a failed gate.
- After two failed attempts, return an honest failure report.
Contribution rules
Every meaningful phase requires one saved prompt, one feature branch, logical commits, tests or evaluation evidence, one draft pull request, and human review before merge.
Use logical commits appropriate to the phase, such as:
- prompt and specification
- tests and fixtures
- implementation
- evaluation evidence
- documentation
Never push directly to main. Never merge automatically.
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.

