dartmouth-skills
AGENTS.md0 starsChanged 34 days agoBSD-3-ClauseView on GitHub
Rules for agents adding, editing, or reviewing skills in the dartmouth-skills repo.
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 32,671 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 repository6 other files it gives its agents.
- dartmouth-brandingSkill
- dartmouth-chatSkill
- decision-modelsSkill
- langchain-dartmouthSkill
- presentation-designSkill
- research-container-deploymentSkill
The file
We split this AGENTS.md into the rules it gives an agent
6 rules. About 1 is common, 1 is distinctive, and 4 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.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 30
Only a human review adds verified: { by: human:<netid>, at: <timestamp> } and sets status: stable.
- Distinctive: line 31
All timestamps are ISO 8601 with an explicit UTC offset.
- Very specific: line 33
Every sources entry has an id and a resource.
- Very specific: line 38
When a file is past stale_after, re-check its sources before trusting it.
- Common: line 40
Public information only: No internal infrastructure details, secrets, keys, or non-public vendor terms.
- Very specific: line 41
Record every change: Add an entry to skills/log.md (newest first, ## YYYY-MM-DD).
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 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
Rules (in priority order)
- One skill per directory:
skills/<name>/SKILL.md.namemust equal<name>: lowercase letters, digits and hyphens, at most 64 characters. descriptiontells agents when to use the skill. Say what it does and when to use it, in one line under 1024 characters. It also serves as the OKF description.- Every
.mdneeds OKF frontmatter with atype. The only exceptions areindex.md,log.mdand the rootREADME.md(kept clean for GitHub). Use:Agent SkillforSKILL.mdReferenceforreferences/*.mdAgent Instructionsfor repo docs such as this file
- Provenance and trust fields:
generated: { by: <actor>, at: <timestamp> }.byis required. Actors are<agent>/<model>for agents,human:<netid>for people,process:<id>for automation.- New or agent-edited content is
status: draftand has noverifiedkey (no key means unverified). - Only a human review adds
verified: { by: human:<netid>, at: <timestamp> }and setsstatus: stable.verifiedis a mapping or a list of such mappings, never a plain name, and thehuman:prefix is what marks content as human-reviewed. - All timestamps are ISO 8601 with an explicit UTC offset.
- Sources:
- Every
sourcesentry has anidand aresource. - Cite key claims with footnotes keyed by that
id([^rc-faq]). - Set
last_modifiedto the source page's own "last modified" date. Omit it if the page shows none.
- Every
- Freshness:
stale_afteris an absolute timestamp. Set it no more than 3 months out for content that names models or prices.- When a file is past
stale_after, re-check its sources before trusting it.
- Keep the
SKILL.mdbody lean (target under 300 lines, excluding frontmatter). Move long configs and walkthroughs intoreferences/and link to them by relative path. - Public information only. No internal infrastructure details, secrets, keys, or non-public vendor terms. Skills dealing with internal APIs or data belong in the Dartmouth-only sibling repo https://github.com/dartmouth/dartmouth-internal-skills, not here.
- Record every change. Add an entry to
skills/log.md(newest first,## YYYY-MM-DD). Updateskills/index.mdwhen adding a skill.
Validate
uv run --with pyyaml python3 scripts/validate.pyThe script needs PyYAML, so uv run --with pyyaml avoids installing it globally.
# alternative if PyYAML is already installed
python3 scripts/validate.pyDoes 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.

