skills-routing
Skill3 starsChanged 10 days agoMITView on GitHub
Select and compose skills when coding-task ownership is unclear or skill-selection guidance is requested. Also resolve open choices about simplification, implementation language, ad hoc tools, or verification methods, and guide session and recovery boundaries. Established workflows, settled methods, and routine commands proceed directly.
In this repository34 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- analyze-projectSkill
- api-contract-strategySkill
- architecture-patternsSkill
- close-changeSkill
- design-changeSkill
- development-standardsSkill
- docker-multiarch-buildSkill
- error-patternsSkill
- go-guidelinesSkill
- herdr-implementSkill
- implement-changeSkill
- logging-standardsSkill
- lua-guidelinesSkill
- organize-docsSkill
- output-stylesSkill
- plan-changeSkill
- powershell-guidelinesSkill
- python-guidelinesSkill
- review-changeSkill
- review-designSkill
- review-implementationSkill
- review-planSkill
- routeros-scripting-guidelinesSkill
- runtime-diagnosisSkill
- security-guardrailsSkill
- shell-guidelinesSkill
- skill-authoringSkill
- skill-minerSkill
- smart-commitSkill
- sops-age-guardrailsSkill
- sync-truthSkill
- testing-strategySkill
The file
Files
Skills Routing
Resolve the question that is actually open. A request that explicitly names or confidently matches another skill goes directly to it; this skill is a method library and a tie-breaker, not a mandatory entry, lifecycle, or runtime dispatcher. Settled language and verification choices, routine commands, and ordinary cleanup stay with the current task.
Read Only The Relevant Branch
A workflow may read a method reference directly without reselecting its primary skill.
| Open question | Read |
| Which skill owns an ambiguous request, or how should skills compose? | Skill Selection and Scope Contracts. |
| What unnecessary responsibility can disappear from code, tests, documents, or supporting mechanisms, or should a bespoke mechanism be replaced? | Simplification. |
| Which language fits new persisted code or an authorized migration? | Language Selection. |
| Which ad hoc CLI or command composition fits a non-trivial task? | Tool Selection; nested interpreters and mutation safety in Ad Hoc Command Composition. |
| What executable evidence would establish the required outcome? | Oracle Selection. |
| How should recalled facts, compaction, or a handoff preserve the current task? | Memory And Recovery; Phase Boundaries when choosing context transfer after a completed phase. |
| Does an ambiguity need clarification, or how does an explicit preference apply? | references/goal-alignment.md in design-change. |
Keep The Current Owner And Authority
Reply composition follows output-styles; method references contribute decisions and evidence without an extra phase or approval gate. A simplification audit stays read-only, and bounded authorized application continues through implement-change. Recovery reconciles actual changes, the current goal and approval baseline, unadjudicated reports, and any still-valid executor; it neither restarts skill selection nor invents authority, and raw sessions are not a way around workspace or permission checks.
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.