optimize-agent-docs
Skill0 starsChanged 14 days agoMITView on GitHub
Create or revise agent-facing rules, skills, instructions, and execution contracts by retaining decisions unavailable from code or help. Use for agent-facing portions of mixed documents, not human manuals.
In this repository6 other files it gives its agents.
- coordinate-repository-versionsSkill
- domain-modelingSkill
- grillingSkill
- grill-with-docsSkill
- human-handoffSkill
- optimize-human-docsSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
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
- Distinctive: line 15
If a requested new guide would only repeat supplied help, return that finding and the existing help entry point rather than manufacturing a …
- Distinctive: line 19
Keep the execution purpose, permission boundaries, environment selection, unrecoverable current state, and acceptance conditions.
- Distinctive: line 32
Do not add warnings or recurrence procedures for easily recoverable problems or problems whose preconditions do not hold on the relevant …
- Distinctive: line 37
If a necessary protection can be implemented but implementation is outside the request, return that concrete remaining change to the …
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill 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
Optimize agent documentation
Identify the requested outcome and authorized edit scope. Preserve the document's language. Inspect only relevant code, help, and existing authoritative documents needed to decide what the agent cannot otherwise recover. Subtract explanations already available there; if nothing necessary remains, omit the new document or remove the redundant one within the authorized scope. Add a short index to a concrete entry point only when it helps the agent find needed information. If a requested new guide would only repeat supplied help, return that finding and the existing help entry point rather than manufacturing a document to fill the requested shape; respect an explicit requirement to retain a separate copy.
Keep the execution purpose, permission boundaries, environment selection, unrecoverable current state, and acceptance conditions. Do not replace important information unavailable from code or help with a reference the executing agent cannot retrieve. Keep one editable authority; moving the same instructions to another document is not a reduction.
For a necessary defense against serious failure, first consider enforcement in the ordinary code path or platform. An optional helper, a test, or an instruction to remember a check is not enforcement in that path. Preserve existing defenses: assess the consequences without them before using current safety as grounds to remove them. Lack of an observed failure does not establish that its preconditions are impossible.
Do not add warnings or recurrence procedures for easily recoverable problems or problems whose preconditions do not hold on the relevant path. Include human monitoring, approval waits, and repeated work in recovery cost. This does not cancel explicit requirements, necessary data protection, or authorization limits.
If a necessary protection can be implemented but implementation is outside the request, return that concrete remaining change to the designated task or issue; do not implement it without authorization or claim completion by relocating its documentation. Retain the judgment that cannot be enforced in code.
Return a focused edit, deletion, or no change. Do not rewrite healthy sections or create an audit report, new ledger, maintenance workflow, or numerical reduction target. The result is the useful remainder, not a record of the optimization.
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.

