agentleFS
igovetigovet/cortex-orch-acp/skills/skill-management/SKILL.md

skill-management

Skill1 starsChanged 59 days agoMITView on GitHub

Activate and scope explicitly requested skill work, or propose a skill from repeated retained cases.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 5

    Distinctive

    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
  1. Very specific:

    Keep user intent separate from runtime authority.

    line 14
  2. Distinctive:

    Use the advertised proposal/question operation with retained case evidence.

    line 32
  3. Distinctive:

    Follow its live schema and approval semantics.

    line 33
  4. Distinctive:

    After approval, the next planning turn carries the /skill-management request the approval stands for.

    line 37
  5. Distinctive:

    After a decline or an unclear answer, leave the project unchanged for that skill and do not propose it again.

    line 38
  6. Distinctive:

    If the runtime refuses a proposal, do not work around it by asking in prose or planning the skill anyway.

    line 40

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Skill management

Activation and ownership

Apply this workflow only to a user request delivered through the exact /skill-management <request> command or to a skill proposal the user approved through the runtime question (see Recurring opportunities). Do not infer authorization from ordinary prose, a recurring task, existing skill files, or a worker's recommendation. If the user invokes the command without a request, ask them to provide one.

The coordinator reads this skill, interprets the request, and proposes a bounded plan through Cortex's advertised planning tools. The runtime owns dispatch, scopes, permissions, user confirmations, independent review, verification, and completion. A skill instruction grants no tool authority; follow the current advertised schemas and use an existing worker role that fits the content.

Keep user intent separate from runtime authority. /skill-management authorizes only the requested skill work, not additional paths or outside-project effects.

Bounded assignment

Choose an existing writing role suited to the skill's content. Assign the requested create, revise or import work and require skill-authoring before drafting or adopting files. Workers own inspection, provenance, resource validation, edits, tests and the guarded Claude link; the coordinator assesses their retained evidence. Require exact changed paths, import identity/license where applicable, actual checks and unresolved link or engine-loading evidence. Do not claim engine use without observation.

Recurring opportunities

When retained reports show the same procedure repeated in separate cases, you may propose one reusable project skill. Do not create it automatically, and never propose one merely because skill files or directories exist in the project. Propose only while planning or reconciling, at a point where the question does not interrupt running workers (for example after verified work, instead of synthesis), and at most once per task.

  1. In your message, explain the repeated cases, what the skill would do, and its benefit, in the language selected by the runtime's per-turn directive.
  2. Use the advertised proposal/question operation with retained case evidence. Follow its live schema and approval semantics; do not invent approval labels.
  3. The runtime validates the evidence and current skill state, presents its evidence list and records the user's decision. An arbitrary first option never approves. A refusal or unclear answer grants no authority.
  4. After approval, the next planning turn carries the /skill-management request the approval stands for. Follow this workflow exactly as for the explicit command.
  5. After a decline or an unclear answer, leave the project unchanged for that skill and do not propose it again. The user can still run /skill-management <request>.

If the runtime refuses a proposal, do not work around it by asking in prose or planning the skill anyway.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.