delegation-routing
Skill2 starsChanged 8 months agoApache-2.0View on GitHub
Encodes the Delegation Decision Gate routing table for executive-tier agents: which specialist agent handles each task domain, and when an agent may act directly instead of delegating. USE FOR: deciding which agent to delegate a phase to before execution begins; verifying a planned direct action is on the "act directly" list; onboarding a new executive-tier agent's delegation posture. DO NOT USE FOR: orchestrating the per-phase checkpoint sequence (use phase-gate-sequence); authoring agent files (use agent-file-authoring); session start/close lifecycle (use session-management).
In this repository27 other files it gives its agents.
- adopt-test-2/AGENTS.mdAGENTS.md
- adopt-test-3/AGENTS.mdAGENTS.md
- adopt-test/AGENTS.mdAGENTS.md
- AGENTS.mdAGENTS.md
- {{cookiecutter.project_slug}}/AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- .github/agents/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- agent-file-authoringSkill
- conventional-commitSkill
- corpus-sweepSkill
- deep-research-sprintSkill
- phase-gate-sequenceSkill
- prompt-archaeologySkill
- pr-review-replySkill
- pr-review-triageSkill
- rate-limit-resilienceSkill
- research-epic-planningSkill
- secondary-research-sprintSkill
- session-managementSkill
- session-retrospectiveSkill
- skill-authoringSkill
- source-cachingSkill
- sprint-planningSkill
- transcript-extractionSkill
- validate-before-commitSkill
- workplan-scaffoldSkill
The file
Files
Delegation Routing
This skill enacts the Algorithms-Before-Tokens axiom from MANIFESTO.md: a static routing table replaces per-session re-derivation of which agent handles which domain, eliminating redundant token burn and preventing scope drift into direct execution.
Beliefs & Context
- Governing axiom: Axiom 2 — Algorithms Before Tokens — deterministic lookup over interactive re-derivation
- GitHub issue: #79 — Skills as Decision Codifiers
- Agents that use this skill: Executive Orchestrator, Executive Researcher, Executive Planner
- Foundation docs:
AGENTS.md— guiding constraints; Minimal Posture, Programmatic-First, delegation-firstexecutive-orchestrator.agent.md— canonical Delegation Decision Gate (lines 137–165)
Workflow
Step 1 — Identify the task domain from the incoming work item or phase description.
Step 2 — Consult the routing table:
| Task domain | Delegate to |
| Research, source gathering | Executive Researcher → Research Scout fleet |
| Documentation writing / editing | Executive Docs |
| Scripting, automation design | Executive Scripter, Executive Automator |
| Fleet agent authoring / audit | Executive Fleet |
| Release coordination, versioning | Release Manager |
| Issue triage, labels, milestones | Issue Triage, Executive PM |
| CI health, test coverage gaps | CI Monitor, Test Coordinator |
| Environment / dependency audit | Env Validator |
| Security threat modelling | Security Researcher |
| Docs compliance audit | Docs Linter |
| Model / cost optimisation | LLM Cost Optimizer |
| Community health, DevRel | Community Pulse, DevRel Strategist |
Step 3 — Verify it is not on the "Act directly" list. An agent may act directly only for:
- Reading files to confirm a deliverable exists
- Running
git status,git log --oneline,gh pr view,gh issue view - Writing scratchpad entries and workplan status updates
- Orchestrator: Running
git add/commit/pushafter a subagent returns (terminal operations) - Running
prune_scratchpad.pyor the pre-compact sequence
If the work does not appear in the "Act directly" list, delegate it.
Step 4 — Form a narrow delegation prompt. Outbound prompts must be task-scoped, not bulk session context. Inbound returns must be ≤ 2,000 tokens (Focus-on-Descent / Compression-on-Ascent — see AGENTS.md § Agent Communication).
Completion Criteria
An agent has applied this skill correctly when:
- The task domain matches exactly one row in the routing table, or is confirmed to be on the "Act directly" list.
- No substantive domain work was performed directly when a specialist agent exists for it.
- The outbound delegation prompt is narrow and task-scoped.
- Phases are delegated one at a time — no batching.
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.

