--- name: aoa-summon description: Delegate one bounded child route through quest-passport law, local-first Codex execution defaults, hard gates for progression, self-agent, and stress posture, and governed return into reviewed closeout. Use when a parent route already has a real anchor and named outputs and the honest next move is a narrower child reviewer, evaluator, verifier, or leaf helper rather than a larger orchestration layer. Do not use when the route lacks an anchor, the outputs are unnamed, the work is `d3+` and still needs a split, or remote delegation is being used to bypass proof, approval, or closeout posture. license: Apache-2.0 compatibility: Designed for Codex or similar coding agents with repository file access and an interactive shell. Network access is optional and only needed when repository validation or referenced workflows require it. metadata: aoa_scope: core aoa_status: scaffold aoa_invocation_mode: explicit-only aoa_source_skill_path: skills/aoa-summon/SKILL.md aoa_source_repo: 8Dionysus/aoa-skills aoa_technique_dependencies: AOA-T-PENDING-SUMMON-LANE-SELECTION,AOA-T-PENDING-SUMMON-RETURN-GOVERNANCE aoa_portable_profile: codex-facing-wave-3 --- # aoa-summon ## Intent Use this skill to decide whether and how to launch one bounded child route from an already-anchored parent route. The goal is not delegation theater. The goal is to preserve quest-passport law, Codex-first local execution, stress narrowing, progression and self-agent gates, return planning, checkpoint-aware reviewed closeout, and owner-local publication mapping. ## Trigger boundary Use this skill when: - a parent route already has a real anchor and named expected outputs - a narrower child actor would help through bounded review, evaluation, leaf implementation, or local verification - the child result must map back into return, closeout, and owner-publication surfaces - local Codex execution is the honest default unless a separate execution surface is actually required Do not use this skill when: - the parent route still lacks a real anchor or named outputs - the route is `d3+` and still needs a split before any child launch - delegation would widen authority, hide approval posture, or bypass a human gate - the task is asking for a whole orchestration mesh rather than one bounded child route - remote transport is being requested for prestige rather than necessity ## Inputs - a quest passport with difficulty, risk, control mode, and delegate tier - a summon request with desired role or child target, transport preference, one parent anchor (`route_anchor`, `parent_task_id`, or `session_ref`), and named expected outputs - named expected outputs for the child route - optional reviewed artifact path, stress bundle ref, checkpoint note ref, Codex trace ref, progression overlay ref, self-agent checkpoint ref, and audit refs ## Outputs - one summon decision with allowed or blocked posture - one chosen lane such as `codex_local_leaf`, `codex_local_reviewed`, `remote_reviewed`, `split_required`, or `human_gate` - `execution_surface`, `cohort_pattern`, `reason_codes`, and `blocked_actions` - optional `codex_local_target` - `return_plan`, `checkpoint_bridge_plan`, `memo_export_plan`, and `owner_publication_plan` - `closeout_required`, `checkpoint_required`, and `progression_required` ## Procedure 1. start from the parent anchor and quest passport, not from raw pressure to delegate 2. default `transport_preference` to `codex_local` when the request leaves it open 3. verify that expected outputs are named before picking a lane 4. classify the lane with difficulty, risk, control mode, requested role, and `references/passport-lane-matrix.v3.md` 5. keep low-risk `d0_probe`, `d1_patch`, and bounded `d2_slice` leaf work in `codex_local_leaf` when anchor and outputs are clear 6. keep local reviewer, evaluator, and architect-like narrowing work in `codex_local_reviewed` 7. allow `remote_reviewed` only when a separate endpoint or execution surface is actually required 8. if difficulty is `d3+`, return `split_required` instead of launching child execution 9. if stress posture says `stop_before_mutation`, only allow a narrowing non-mutating child; otherwise gate to human review 10. if `require_progression` is true or the route depends on reviewed unlock posture, require a reviewed progression overlay before summon 11. if the route is self-agent or policy-sensitive, require reviewed self-agent checkpoint posture before summon 12. emit return, checkpoint bridge, memo export, and owner publication plans that stay subordinate to reviewed closeout and owner truth 13. apply `references/no-raw-traces-rule.md` so child traces remain aids rather than proof or memo canon 14. map failed, narrowed, or blocked child states back into return posture instead of letting them disappear ## Contracts - this skill governs one bounded child route; it does not grant hidden orchestration authority - local Codex child targeting is the default first choice, not an afterthought - the SDK-owned E2E fixture at `repo:aoa-sdk/examples/a2a/summon_return_checkpoint_e2e.fixture.json` may be used to check this skill's v3 request/result contract, but it does not move summon law into `aoa-sdk` - stress may narrow or block, but it may never widen authority - remote transport follows the same passport, return, checkpoint, and closeout law as local transport - `d3+` routes split first - progression and self-agent gates are real blockers when required evidence is missing - child traces, memos, and receipts do not replace rereading the reviewed parent artifact - terminal publication stays inside canonical owner families, and memo export stays reviewed and candidate-oriented - failed, narrowed, or blocked child results still map into an explicit return surface ## Risks and anti-patterns - treating delegation as progress without anchor or output clarity - using a remote child route to bypass local proof, approval, or closeout posture - allowing child scope to widen beyond the parent passport - reading child traces as proof authority or memo canon - skipping the required split for `d3+` work - publishing terminal meaning into derived layers before owner-local landing ## Verification - confirm the passport and parent anchor exist - confirm expected outputs are named - confirm the chosen lane matches difficulty, risk, control mode, and requested role - confirm any E2E fixture check validates `summon_request` and `summon_result` against `references/summon-request-v3.schema.json` and `references/summon-result-v3.schema.json` - confirm `d3+` routes split before summon - confirm stress only narrows or blocks - confirm progression and self-agent gates are not bypassed - confirm return, checkpoint bridge, and publication plans stay explicit for nonterminal or failed child outcomes - confirm terminal publication stays owner-local and memo export stays reviewed and candidate-oriented - confirm traces stay subordinate to reviewed closeout and parent reread ## Technique traceability Manifest-backed techniques: - AOA-T-PENDING-SUMMON-LANE-SELECTION from `8Dionysus/aoa-techniques` at `TBD` using path `TBD` and sections: Intent, Inputs, Outputs, Core procedure, Contracts, Validation - AOA-T-PENDING-SUMMON-RETURN-GOVERNANCE from `8Dionysus/aoa-techniques` at `TBD` using path `TBD` and sections: Intent, Outputs, Contracts, Risks, Validation ## Adaptation points - project overlays may add local child-role labels, control modes, or transport labels - project overlays may map local reviewed-closeout or checkpoint families onto the core summon result without weakening the gates - project overlays may specialize owner publication families or memo candidate classes while keeping the parent return route explicit