skills / instructions
AGENTS.md0 starsChanged 2 days agoMITView on GitHub
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 31,439 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 repository22 other files it gives its agents.
The file
We split this AGENTS.md into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 4
When it is unclear which skill should own the result, consult the active Mirket plugin's skills/execute/references/Routing.md.
- Distinctive: line 6
Read the project's instructions and current state before changing it.
- Distinctive: line 9
Use focused checks and required project checks.
≈ “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.
Files
Mirket working instructions
- On each request, identify and use the smallest relevant set of skills from the active Mirket plugin before substantive work. Apply Execute throughout: take the simplest complete path and report what matters. Use Author for general writing and issue triage, Market for marketing messaging and claims, and Ship for product requirements. If Mirket is unavailable, say so, follow these defaults, and continue authorized work.
- When it is unclear which skill should own the result, consult the active Mirket plugin's
skills/execute/references/Routing.md. Add support at the start or during work when it can answer a distinct question, fill an evidence gap, or guide a decision. - Use Audit for security and privacy review and Architect for design-time threat modeling. The implementer owns authorized hardening. The task owner follows
skills/execute/references/VersionedDistribution.mdfor release preparation and authorized publication. Use Migrate for persistent-state transitions and Deploy for environment rollout, service health, incidents, and recovery. Honor existing authority; preparation, publication and production activation are separate scopes. - Read the project's instructions and current state before changing it. Respect existing conventions and user changes. Prefer the smallest complete solution; challenge a poor tool fit with evidence and a better option.
- Plan substantial implementation with Plan. Ground decisions in evidence, keep product and business choices with the human, and record the agreed outcome and checks. Keep routine work direct and read-only reviews free of execution plans.
- Use
.work/at the current Git worktree root or the selected non-Git project root:plansfor planning,decisionsfor private architecture drafts,stashesfor temporary snapshots,recoverfor important recovery copies only,briefsfor standalone handoffs and briefs, andtempfor disposable files. Number new plans and standalone handoffs using the active plugin'sskills/execute/references/Workspace.md. Record settled architecture decisions in the project's ADR location, creatingdocs/decisions/if none exists. Use designated paths for real deliverables. Before writing to.work/, ensure/.work/is ignored in that repository, with a local Git exclude if appropriate, and check for tracked workspace files. Keep it out of the index unless the human explicitly names an exception. - Use focused checks and required project checks. Inspect rendered changes before claiming visual quality. Apply Clean before delivery and before committing; preserve unrelated work and important records.
- Lead with the result, cite useful evidence, and state material limits. Report only checks and actions that actually happened.
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.

