sa-aox
Skill0 starsChanged 4 months agoMITView on GitHub
SA-AOX (Stand-Alone IdeaFirst Orchestrator eXplorer) — cross-model 평가가 미가용인 단일 모델 런타임에서 SDX/TCX/IDX 산출물을 받아 SA-ICX와 SA-EVX를 순차 실행하고 standalone final summary와 production promotion package를 만드는 오케스트레이터. AOX production을 대체하지 않고 .aox production run을 completed로 위장하지 않으며 .sa-aox/에만 산출한다. Triggers: sa-aox, standalone aox, stand-alone aox, 단독 AOX, 독립 AOX, standalone full cycle, SA IdeaFirst
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
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
- Very specific: line 230
If CIX import from .sa-icx is not implemented, state manual handoff:
- Very specific: line 239
Use inline PGF for a single run.
- Very specific: line 243
If a selected SA-AOX winner is implemented as a project, MVP, repo, or durable PGF track, SA-AOX must record it in …
- Distinctive: line 272
Do not record candidates that were only generated, evaluated, or discussed.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.
Files
SA-AOX (Stand-Alone IdeaFirst Orchestrator eXplorer) v0.2
SA-AOX는 cross-model 평가가 미가용인 단일 모델 런타임(Claude Code, Codex 등)에서 IdeaFirst exploratory loop를 끝까지 닫는 standalone 오케스트레이터다.
런타임이 subagent 병렬 파견을 지원하면 SA-ICX/SA-EVX의 페르소나 작업을 병렬 subagent로 실행하는 것이 기본이다(PGF discover 패턴, pgf/agent-protocol.md의 PG TaskSpec 사용). 단, subagent는 같은 모델이므로 결과는 여전히 single-model이며 standalone boundary는 변하지 않는다.
Completion Definition
SA-AOX is complete when the standalone pipeline runs by composing official skills only. SA-AOX must not become a duplicate implementation of SA-ICX or SA-EVX; it must orchestrate them, verify their artifacts, enforce consumed-idea exclusion, and write the standalone wrap-up.
sa_aox_completion_criteria:
source_of_truth:
- "skills/sa-aox/SKILL.md"
- "skills/sa-icx/SKILL.md"
- "skills/sa-evx/SKILL.md"
- "skills/pg/SKILL.md"
- "skills/pgf/SKILL.md"
- ".idea-ledger/consumed_ideas.yaml"
excluded_from_pipeline_contract:
- "aox_agents.md"
complete_when:
- "SA-AOX can run Stage0..Stage5 through skill contracts only"
- "SA-ICX candidate_pool is produced or reused through SA-ICX contract"
- "SA-EVX final_idea and dual winner are produced through SA-EVX contract"
- "consumed winners are excluded before acceptance"
- "implemented winners are recorded in the consumed idea ledger"
- "all outputs stay under .sa-icx, .sa-evx, and .sa-aox"
incomplete_when:
- "SA-AOX depends on aox_agents.md or other handoff notes as policy"
- "SA-AOX bypasses SA-ICX/SA-EVX contracts by inventing untracked outputs"
- "SA-AOX repeats an idea already consumed into a project"Boundary
SA-AOX는 AOX production 결과가 아니다.
result_class: "standalone_single_runtime"
cross_model_certified: false
production_aox_equivalent: false
may_write_aox_completed_summary: false
may_write_cix_latest: false
may_write_evx_latest: falseSA-AOX가 만드는 것은 최종 인증 결과가 아니라:
- standalone final idea
- standalone summary
- production promotion package
- CIX/EVX/AOX로 승격하기 위한 trace
Pipeline
SA_AOX_Full
Stage0_Init
create .sa-aox/{run_id}/status.json
probe file_io and PGF personas
record cross_model_capability but do not block standalone flow
load .idea-ledger/consumed_ideas.yaml if present
Stage1_UpstreamCheck
verify .sdx/catalog/index.yaml
verify .tcx/latest/
verify .idx/latest/insight_layered_traced.yaml
verify consumed idea ledger is readable when it exists
Stage2_SA_ICX
invoke or execute SA-ICX forge
require consumed ideas to be excluded before candidate_pool.yaml is finalized
output .sa-icx/rounds/{sa_icx_round_id}/candidate_pool.yaml
Stage3_SA_EVX
invoke or execute SA-EVX evaluate
reject any consensus_winner or innovation_winner that appears in consumed_ideas.yaml
output .sa-evx/rounds/{sa_evx_round_id}/final_idea.md
Stage4_WrapUp
write .sa-aox/{run_id}/summary.md
write .sa-aox/{run_id}/PROMOTE_TO_PRODUCTION.md
update .sa-aox/index.yaml
Stage5_ConsumptionRecord
when standalone winner is converted into a concrete project folder or MVP
record it with the production ledger tool (never hand-roll the schema):
python scripts/explore/aox_full.py --project-root . --mode wrapup --record-consumed
--idea-id {winner id} --idea-title "{winner title}"
--project-name {Name} --project-path {name}
--repo-url {url} --aliases {Name},<alt> --semantic-family <slug>
note: SA winner ids are absent from .cix/latest/idea_pool.yaml, so --idea-title is REQUIRED
then validate: python scripts/explore/ledger_lint.py --project-root .
do not record merely explored candidates that were not selected or implemented
verify future SA-ICX/SA-EVX runs exclude the consumed winnerInputs
required:
sdx_catalog_index: ".sdx/catalog/index.yaml"
tcx_latest: ".tcx/latest/manifest.yaml"
idx_latest: ".idx/latest/insight_layered_traced.yaml"
personas: "skills/pgf/discovery/personas.json"
consumed_ideas_ledger: ".idea-ledger/consumed_ideas.yaml"
optional:
existing_sa_icx_round: ".sa-icx/rounds/{round_id}"
existing_sa_evx_round: ".sa-evx/rounds/{round_id}"
category_map_builder: "skills/cixx/scripts/build_category_map.py" # SA-ICX 조향용
ledger_lint: "scripts/explore/ledger_lint.py"
record_consumed_tool: "scripts/explore/aox_full.py (--record-consumed --idea-id --idea-title)"Outputs
output_root: ".sa-aox"
run_id_format: "SA-AOX-{YYYYMMDD}-{NNN}"
files:
- status.json
- summary.md
- standalone_kpis.yaml
- PROMOTE_TO_PRODUCTION.md
- run_manifest.yaml
- optional consumed idea ledger update when a winner is implementedStatus Contract
status:
run_id: "SA-AOX-{YYYYMMDD}-{NNN}"
mode: "full | resume | dry-run"
result_class: "standalone_single_runtime"
stages:
0_init: "completed | failed"
1_upstream_check: "completed | failed"
2_sa_icx: "completed | failed | skipped_reused"
3_sa_evx: "completed | failed | skipped_reused"
4_wrapup: "completed | failed"
sub_round_ids:
sa_icx_round_id: "SA-ICX-{YYYYMMDD}-{NNN}"
sa_evx_round_id: "SA-EVX-{YYYYMMDD}-{NNN}"
consumed_idea_gate:
ledger_path: ".idea-ledger/consumed_ideas.yaml"
excluded_count: integer
exclusion_report: ".idea-ledger/exclusion_report_latest.md"
winner_reuse_detected: false
consumption_record:
required_when_project_created: true
ledger_path: ".idea-ledger/consumed_ideas.yaml"
record_via: "aox_full.py --record-consumed --idea-id --idea-title (직접 append 금지)"
lint_after_record: "scripts/explore/ledger_lint.py errors=0"
reuse_policy: "exclude_same_or_derivative"
recorded_idea_id: "string | null"
recorded_project_path: "string | null"
production_boundary:
cross_model_certified: false
cix_promotion_required: true
evx_production_required: true
aox_production_wrapup_required: trueStandalone KPIs
SA-AOX may record analogous exploratory metrics, but they must not be confused with AOX v1.3.1 production KPIs.
standalone_kpis:
candidate_diversity_proxy: number
layer_balance: {L6_Gap, L7_Tension, L9_Counterfactual, L10_Generative}
persona_vote_breadth: number
consensus_vs_innovation_split: boolean
production_certification_status: "not_certified"Forbidden KPI language:
novelty baseline failure rateunless cross-model CIX has runsurprise pass rateunless cross-model CIX has runAOX KPI passed
Summary Contract
summary.md must include:
# SA-AOX Standalone Summary
This is a standalone IdeaFirst result, not CIX/EVX/AOX production-certified output.
Cross-model CIX v1.5.1 promotion is required before production use.
## Outputs
- SA-ICX candidate pool
- SA-EVX final idea
- SA-EVX dual winner block
## Boundary
- cross_model_certified: false
- may_use_for_exploration: true
- may_use_for_final_certification: false
## Production promotionPromotion Package
PROMOTE_TO_PRODUCTION.md must give the exact next production path:
1. Run CIX v1.5.1 cross_model surprise_validation using SA-ICX candidates as evidence.
2. Emit completed .cix/latest only after cross_model validation.
3. Run EVX v1.1 production evaluate.
4. Run AOX v1.3.1 wrap-up and collect production 5 KPIs.If CIX import from .sa-icx is not implemented, state manual handoff:
Use .sa-icx/rounds/{id}/candidate_pool.yaml and raw_seed_ideas.yaml as candidate evidence.
Do not copy them into .cix/latest without CIX v1.5.1 validation.PGF Execution Rule
Use inline PGF for a single run. Create durable .pgf/ workplans only when changing this skill, running a multi-turn standalone campaign, or handing off to another environment.
Consumed Idea Recording
If a selected SA-AOX winner is implemented as a project, MVP, repo, or durable PGF track, SA-AOX must record it in .idea-ledger/consumed_ideas.yaml before the task is considered complete.
Record via the production tool — the ledger schema is owned by AOX and enforced by ledger_lint.py; SA-AOX must not hand-roll entries:
python scripts/explore/aox_full.py --project-root . --mode wrapup --record-consumed \
--idea-id {winner id} --idea-title "{winner title}" \
--project-name {Name} --project-path {name} --repo-url {url} \
--aliases {Name},<alt1> --semantic-family <slug>
python scripts/explore/ledger_lint.py --project-root .The resulting entry follows the production schema (reference only — do not write by hand):
- idea_id: "winner id" # SA round-local id
title: "winner title"
normalized_title: "kebab-case of title"
aliases: ["{Name}", "<alt1>"]
semantic_family: "single-slug-string" # string, not list
source_chain: {cix: "...", idx: "...", tcx: "...", evx: "...", sdx_catalog: "..."}
consumed_at_utc: "ISO-8601"
implementations:
- {project_name: "{Name}", project_path: "{name}", repo_url: "{url}"}
reuse_policy: "exclude_same_or_derivative"SA round ids (SA-AOX/SA-ICX/SA-EVX-...)는 ledger 스키마 밖이므로 .sa-aox/{run_id}/run_manifest.yaml에 기록하고, ledger에는 위 정본 키만 남긴다.
Do not record candidates that were only generated, evaluated, or discussed. Record only after the idea has been consumed into concrete work.
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.

