feature-loop
Skill6 starsChanged 3 months agoMITView on GitHub
Coding agent for the project-level harness — one feature per session, must verify, must be mergeable.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- ruyoutor/agentic-archflowtheir own version, first seen Oct 13, 2026
- dorian-akhs/skillstheir own version, first seen Sep 1, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- dorian-akhs/skills★ 0same name
- ruyoutor/agentic-archflow★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository346 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .claude/skills/deeppapernote/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .claude/skills/deeppapernote/CLAUDE.mdCLAUDE.md
- docs/llms-full.txtllms.txt
- docs/llms.txtllms.txt
- llms.txtllms.txt
- academic-cv-builderSkill
- agenticx-a2a-connectorSkill
- agenticx-agent-builderSkill
- agenticx-automation-crontaskSkill
- agenticx-deployerSkill
- agenticx-memory-architectSkill
- agenticx-quickstartSkill
- agenticx-skill-managerSkill
- agenticx-tool-creatorSkill
- agenticx-workflow-designerSkill
- ai-slop-detectorSkill
- alert-managerSkill
- algorithmic-artSkill
- api-gatewaySkill
- backlink-analyzerSkill
- brand-guidelinesSkill
- canvas-designSkill
- career-changer-translatorSkill
- Codex-apiSkill
- Codex-skill-rustSkill
- claw-releaseSkill
- clawsec-clawhub-checkerSkill
- clawsec-feedSkill
- clawsec-nanoclawSkill
- clawsec-scannerSkill
- clawsec-suiteSkill
- clawtributorSkill
- cn-statsSkill
- codebase-documenterSkill
- code-dev-workflowSkill
- competitor-analysisSkill
- content-experimentation-best-practicesSkill
- content-gap-analysisSkill
- content-modeling-best-practicesSkill
- content-quality-auditorSkill
- content-refresherSkill
- context-mode-opsSkill
- context-modeSkill
- cover-letter-generatorSkill
- creative-portfolio-resumeSkill
- crypto-reportSkill
- ctx-doctorSkill
- ctx-insightSkill
- ctx-purgeSkill
- ctx-statsSkill
- ctx-upgradeSkill
- deep-cleanSkill
- doc-coauthoringSkill
- docxSkill
- domain-authority-auditorSkill
- edgeoneSkill
- edge-ttsSkill
- entity-optimizerSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 0 are distinctive, and 7 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.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
Use when session_mode == feature_loop and the project is past the Initializer phase (system prompt will say "Coding 阶段").
- Very specific: line 28
Sync from disk: Call project_status first thing every session — it reads feature_list.json / status.json / progress.md and reseeds your …
- Very specific: line 30
If status.active_feature_id is already set, resume that one (the prompt block will tell you).
- Very specific: line 34
Read with file_read slices (start/end line).
- Very specific: line 36
Run the gate: Call verify_run feature_id=<id>.
- Very specific: line 46
Promote to committed: Call feature_complete feature_id=<id> commit_sha=<sha>.
- Very specific: line 51
verify_run timeouts: do not retry blindly. Inspect the log under .agx/project/archive/, fix the root cause, then rerun.
≈ “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
Feature Loop
Use when session_mode == feature_loop and the project is past the Initializer phase (system prompt will say "Coding 阶段").
Iron rules
- One feature per session. Multiple
in_progressfeatures are forbidden by the store; do not bypass. - Always verify before commit.
feature_completerejects features that have not been promoted toverifiedviaverify_run. - Keep the branch mergeable. Each commit must compile, pass
verify.yaml, and not depend on uncommitted local state.
Loop
- Sync from disk. Call
project_statusfirst thing every session — it readsfeature_list.json/status.json/progress.mdand reseeds your understanding from the single source of truth. - Pick a feature.
- If
status.active_feature_idis already set, resume that one (the prompt block will tell you). - Otherwise call
feature_selectwith no arguments to auto-pick the highest-priority pending feature with satisfied dependencies, or passfeature_idwhen the user names one.
- If
- Implement using code_dev phases (Explore → Read → Author):
- Explore with
code_outline,grep, optionalcode_search. - Read with
file_readslices (start/end line). Track files you have read in scratchpad. - Author with
file_writeskeletons first, then section-by-section appends.
- Explore with
- Run the gate. Call
verify_run feature_id=<id>. If any step fails:- Do not call
feature_complete. - Log a
progress_appendwith the failing step name and root cause. - Decide: fix and rerun
verify_run, or escalate to the user with the specific failure.
- Do not call
- Commit via shell.
Capture the resulting sha (git add -A git commit -m "feat(<feature_id>): <one-line summary>"git rev-parse HEAD). - Promote to committed. Call
feature_complete feature_id=<id> commit_sha=<sha>. The store writes an immutable archive snapshot under.agx/project/archive/feature_<id>.json. - Decide next step. Either call
feature_selectfor the next feature in the same session, or stop and tell the user the loop closed cleanly.
Failure modes
- verify_run timeouts: do not retry blindly. Inspect the log under
.agx/project/archive/, fix the root cause, then rerun. - Dependency missed:
feature_selectwill reject features whosedepends_onisn'tcommitted. Either pick a different feature or finish the prerequisite first. - Lost context after window compaction: re-call
project_status. Disk is the source of truth. - Cold start (new machine, fresh session):
project_status+feature_selectis enough to resume. No memory of previous turns is required.
Forbidden
- Editing files inside
.agx/project/archive/(immutable). - Calling
project_initoutside the Initializer phase. - Skipping
verify_runbecause "tests are too slow".
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.

