agentleFS
Sign inStart free
ivanrvpereiraivanrvpereira/.agents/skills/on-demand/suggest-skills/SKILL.md

suggest-skills

Skill19 starsChanged 4 days agoApache-2.0View on GitHub

Suggest which installed skill fits the current step of your work.

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 4 are distinctive, and 3 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.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read the chosen skill's SKILL.md to confirm it fits.

    line 13
  2. Very specific:

    Keep one session from shaping the idea through to-tickets.

    line 30
  3. Distinctive:

    When the user asks whether to continue, start fresh, hand off, delegate, or compact, take the first yes:

    line 58
  4. Distinctive:

    Continue: if the next phase needs this session's reasoning (grilling into planning is the usual case) and it fits.

    line 60
  5. Distinctive:

    Start fresh: if nothing here matters to what comes next.

    line 61
  6. Very specific:

    handoff: if the work moves to another harness, directory, or person.

    line 62
  7. Distinctive:

    Subagent: if the task is tightly scoped and needs no steering.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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.

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

Suggest Skills

Suggest the skill for the user's next step, then stop. Read anything you need; run nothing else. A suggestion is not permission: delegating, committing, or publishing still needs the user's own request.

  1. Work out the goal and where the session stands from the conversation. If one missing fact would change the answer, ask one question.
  2. Pick candidates from the maps below. If nothing fits, look wider: the harness's skill list, rg -H '^description:' ~/.agents/skills/{auto,on-demand}/*/SKILL.md ~/.agents/skills/private/*/*/SKILL.md, and the project's .claude/skills/ or .pi/skills/. Manual-only skills are missing from the harness's list.
  3. Read the chosen skill's SKILL.md to confirm it fits. The installed skill wins over these maps.
  4. Reply with the skill, one sentence on why, what comes after it when that helps, and how to invoke it (/name in Claude Code, /skill:name in Pi, $name in Codex). If a close alternative exists, give the deciding difference in one sentence.

Build a feature

Work down the steps; skip the ones the situation does not need.

StepSkillWhen
Shape the ideagrill-with-docsWorking in a repo; it records GLOSSARY.md and ADRs. Use grill-me outside a repo, or grill-for-unknowns when the risk is unknown unknowns in a complex build.
Settle a design questionprototypeTalk cannot settle how logic should behave or a UI should look. Bring the answer back to the plan.
Planto-spec, then to-ticketsA build too big for one session. Use tech-spec instead for a typed call-stack design document rather than a tracker spec.
BuildimplementOne ticket, or a small task whose scope and test seams are agreed. It runs tdd and code-review itself. Start each ticket in a fresh session.
Build with subagentsimplementation-orchestratorThe user wants a whole approved plan executed by subagents.
ShipprWriting the pull request body.
Look backretroAfter a build, before the session is cleared.

Keep one session from shaping the idea through to-tickets. to-spec, to-tickets, and wayfinder need a configured tracker: if docs/agents/issue-tracker.md is missing, suggest setup-matt-pocock-skills first.

rpi is a separate end-to-end flow (research, design, implement) with its own documents. Suggest it only when the user prefers it over the chain above.

Other situations

SituationSkillThen
Hard bug, flaky test, or slowdowndiagnosing-bugs (sentry first for a Sentry error)retro
Effort too large and foggy to plan in one sessionwayfinderto-spec once the decisions settle
Facts needed from docs or sourcesresearchgrill-with-docs
The answer is in someone else's headto-questionnairegrill-with-docs or to-spec
Find refactor candidatesimprove-codebase-architecture (visual report, then grill one), or architecture-scan (ranked TypeScript shortlist, no design)grill-with-docs on the chosen one
Design one module's interface or seamcodebase-designtdd
Shrink a design or codebasereducing-entropy
Build a behavior test-firsttdd
Review a changecode-reviewpr
Fuzzy project termsdomain-modeling
Record an architecture decisionadr-writing
Steps only a human can dowizard
A message did not landwait-what (re-pitch with the missing context) or bro (plain restatement)
Understand something visuallyshow-me
Learn a concept over several sessionsteach
Write or trim skills and AGENTS.mdwriting-for-agents, or skill-diet to trim
Coordinate several independent workstreamssupervisor, only when the user asks for that role

Between phases

When the user asks whether to continue, start fresh, hand off, delegate, or compact, take the first yes:

  1. Continue if the next phase needs this session's reasoning (grilling into planning is the usual case) and it fits.
  2. Start fresh if nothing here matters to what comes next.
  3. handoff if the work moves to another harness, directory, or person.
  4. Subagent if the task is tightly scoped and needs no steering.
  5. Compact otherwise, saying what the next phase needs kept.

Decide at the end of a phase, not in the middle. Every option except continue replaces the full session with a lossy summary.

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.

Check your setup, free