act-phase
Skill0 starsChanged 52 days agoMITView on GitHub
Act phase of an ArcheFlow run: consolidate findings, route them to Creator or Maker, decide merge / cycle back / stop, write act-feedback.md.
In this repository29 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- .cursor/rules/archeflow.mdcCursor rule
- archeflow-reviewSkill
- archeflow-runSkill
- archeflow-sprintSkill
- autonomous-modeSkill
- check-phaseSkill
- cost-trackingSkill
- custom-archetypesSkill
- dagSkill
- domainsSkill
- git-integrationSkill
- initSkill
- memorySkill
- multi-projectSkill
- presenceSkill
- progressSkill
- replaySkill
- reportSkill
- reviewSkill
- runSkill
- scanSkill
- scoreSkill
- shadow-detectionSkill
- sprintSkill
- statusSkill
- templatesSkill
- using-archeflowSkill
- workflow-designSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 0 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 15
Read this cycle's check-*.md.
- Very specific: line 21
Write .archeflow/artifacts/<run_id>/findings-cycle-<N>.json, an array of
- Very specific: line 62
Write .archeflow/artifacts/<run_id>/act-feedback.md.
- Very specific: line 91
leave findings-cycle-<N>.json and convergence-cycle-<N>.json in place.
≈ “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
Act Phase
Turn the Check output into a decision. Act never edits code: every fix goes through the next
cycle and is reviewed again. The commands around these steps are in archeflow:run, step 4.
Step 1: Consolidate
Read this cycle's check-*.md; findings the evidence gate downgraded read INFO there and count as
INFO. Deduplicate: same file + same category + similar description = one finding with the
higher severity, crediting all sources. Cycle 2+: compare with findings-cycle-<N-1>.json:
resolved (gone), persisting (same id, cycles_open + 1), new (cycles_open: 1); emit
fix.applied for each finding a previous fix resolved.
Write .archeflow/artifacts/<run_id>/findings-cycle-<N>.json, an array of
{"id": "src/auth/handler.ts:security", "file": "src/auth/handler.ts", "line": 48,
"category": "security", "severity": "CRITICAL", "source": ["guardian"],
"description": "Empty string bypasses validation", "routes_to": "creator", "cycles_open": 1}id = <file>:<category> (:<n> for a second distinct finding there). Keep ids stable across
cycles: convergence and the Wiggum Break check compare them.
Step 2: Route
| Source | Category | Routes to |
| Guardian | security, breaking-change, reliability, dependency | Creator |
| Skeptic | design, scalability | Creator |
| Sage | quality, consistency, testing | Maker |
| Trickster | reliability (design flaw) | Creator |
| Trickster | reliability (test gap), testing | Maker |
A fix that changes the approach goes to the Creator; one within the approach to the Maker. When merged sources disagree, the Creator wins.
Step 3: Decide
First match wins. The last two columns are the cycle.boundary values.
| Condition | Decision | exit_condition | decision |
| Wiggum Break (hard, or soft at the end of this step) | stop, keep the branch, report | wiggum_break | stop |
| Same CRITICAL open 2+ consecutive cycles | ask the user how to proceed | escalated | escalate |
fast and Guardian found 2+ CRITICAL | next cycle runs as standard (Skeptic + Sage, no fast path) and stays escalated | escalated | cycle_back |
| 0 CRITICAL and every reviewer APPROVED | merge (archeflow:run, Merge) | approved | merge |
| Findings left, cycles left | cycle back (Step 4) | findings_open | cycle_back |
| Findings left, no cycles left | stop, report open findings, keep the branch | max_cycles | stop |
WARNING and INFO alone never block a merge; list them in the report.
Step 4: Feedback
Write .archeflow/artifacts/<run_id>/act-feedback.md; each section goes only to its agent
(Creator at Plan, Maker at Do). Keep it under ~500 tokens: drop INFO first, then summarise
WARNINGs by theme. An empty Creator section means the next cycle keeps the proposal and starts
at Do.
## Cycle <N> -> Cycle <N+1>
## Creator-Routed Issues
| # | Source | Severity | Category | Location | Issue | Cycles open |
|---|--------|----------|----------|----------|-------|-------------|
## Maker-Routed Issues
| # | Source | Severity | Category | Location | Issue | Cycles open |
|---|--------|----------|----------|----------|-------|-------------|
## Resolved This Cycle
| # | Source | Issue | How resolved |
|---|--------|-------|--------------|Step 5: Archive the cycle
In .archeflow/artifacts/<run_id>/cycle-<N>/:
- copy
plan-*.mdandact-feedback.md(the next cycle reads them from the top level; a new Creator run overwritesplan-creator.md); - move
do-*andcheck-*(integrate writes a new diff; old reviews must not reach the next Check); - leave
findings-cycle-<N>.jsonandconvergence-cycle-<N>.jsonin place.
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.

