agentleFS
AhmiDarrowAhmiDarrow/RemedyAI/skills/self-dev-loop/SKILL.md

self-dev-loop

Skill5 starsChanged 23 days agoApache-2.0View on GitHub

Meta-loop for Remedy developing herself — self-inject, change-safety, unit tests, security gauntlet, product soak, stress, then ship gates. Use when self-dev, hot-inject, improve the codebase, work on RemedyAI monorepo end-to-end, or "run the full quality loop on ourselves".

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read root AGENTS.md for ship gates, installer naming, smoke matrix.

    line 74
  2. Distinctive:

    Prefer durable fixes; keep tree runnable.

    line 86
  3. Very specific:

    Follow project-etiquette exactly: no publish on red CI.

    line 109

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Self-dev loop (Remedy on Remedy)

Goal

One ordered playbook when the user wants Remedy to improve this monorepo while staying a trustworthy partner. Prefer activating sub-skills rather than re-inventing commands.

When to use

  • Working in the RemedyAI repo as the project
  • User says self-dev, dogfood loop, full quality loop, “run gauntlet then soak”
  • Multi-hour sessions that must not skip security or ship discipline

Default sequence

Execute in order. Skip a stage only if the user explicitly narrows scope (e.g. “unit tests only”) or the stage does not apply (docs-only change).

0. Orient        → project-overview / AGENTS.md / git status
1. Hot-inject    → self-inject (draft → test-gate → apply/rollback → restart)
2. Blast radius  → change-safety
3. Implement     → write-tests, refactor-safe as needed
4. Unit gates    → pytest + desktop npm test/build + check_docs
5. Gauntlet      → gauntlet-security (if security surface touched or user asked)
6. Soak          → soak-product (if desktop/API/CUA/stream touched or user asked)
7. Stress        → stress-suite (optional; after soak green or user asked)
8. Ship          → project-etiquette (test → docs → build → commit → CI → publish)

Activate helpers

Never skill_activate every pack (“reload all skills”). That floods context and causes endless tool loops. Rescan disk with:

skill_reload()

Then load one procedure at a time (or let auto-suggest inject for the task):

skill_activate(skill="self-dev-loop")   # this meta playbook
skill_activate(skill="self-inject")
skill_activate(skill="gauntlet-security")
skill_activate(skill="soak-product")
skill_activate(skill="stress-suite")
skill_activate(skill="change-safety")
skill_activate(skill="project-etiquette")

Stage details (short)

0. Orient

git status -sb
git log -5 --oneline

Read root AGENTS.md for ship gates, installer naming, smoke matrix.

1. Hot-inject

  • Draft a change with normal edit tools (Python sidecar or desktop SPA).
  • Gate is tests-only (pytest + npm test). Green → apply + restart sidecar / rebuild SPA; red → roll back and record in the ledger (~/.remedy/self_inject_ledger.jsonl).
  • Full mechanics: docs/SELF_INJECT.md. Never commit red; never leave a broken tree.

2–3. Change + implement

  • Name blast radius (chat, gateway, rails, settings, packaging).
  • Prefer durable fixes; keep tree runnable.
  • Pair with write-tests for behavior changes.

4. Unit gates (minimum always before ship)

uv run pytest -q
cd desktop; npm test; npm run build
uv run python scripts/check_docs.py

5–7. Live quality (when scope needs it)

StageSkillTrigger
Gauntletgauntlet-securityauth, jail, SSRF, host, sanitize
Soaksoak-productCUA, browser, stream, multi-tab, “soak”
Stressstress-suiteconcurrency, multi-provider, “stress/break”

Set REMEDY_API / REMEDY_HOME to the profile under test.

8. Ship

Follow project-etiquette exactly: no publish on red CI; version only for real releases; desktop tag vX.Y.Z for installers.

Reporting template

StageResultEvidence
Hot-injectapplied / rolled back / n/aledger + diff-id
Blast radiussurfaces listed
Unitpass/failcommands
Gauntletpass/fail/skip
Soakpass/fail/skipdocs/_full_product_soak_results.json
Stresspass/fail/skip
Shipcommit / CI / publish

Anti-patterns

  • Stress before unit green
  • Gauntlet against the wrong API port/home
  • Shipping without project-etiquette when user asked for release
  • Dual messenger pollers "for testing"
  • Committing a red self-inject round (never — roll back instead)
  • Claiming self-dev done when only the chat replied and no commands ran

Related skills

  • self-inject · gauntlet-security · soak-product · stress-suite
  • change-safety · project-etiquette · github · session-handoff

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.