Skip to content
Back to skills

Plans Yolo

ASecurity

Implement PR 1 from the README brainstorm. Publish one discoverable operator skill, simplify the README, and guard its distribution in CI. The landing repository remains a later PR.

  • 844 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 21, 2026
ai-agentsgodockertestinggitapidatabasedocumentation

Works with

  • api

Security analysis

A92/100
  • mediumUses curl or wget to download content

Pro scans all 12 files and shows the line behind each finding

Scanned September 21, 2026

npx -y skills add desplega-ai/agent-swarm --skill plans-yolo --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Plans Yolo?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Plans Yolo
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/desplega-ai-plans-yolo/badge)](https://www.skillsdirectory.com/skills/desplega-ai-plans-yolo)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
date: 2026-09-08T15:00:00Z
topic: "Public operator skill and README simplification"
status: done
---

# Public operator skill and README simplification

## Goal

Implement PR 1 from the README brainstorm. Publish one discoverable operator skill, simplify the README, and guard its distribution in CI.
The landing repository remains a later PR.

## Decisions

- Use three phases within public documentation and skill distribution (assumed).
- Follow the explicit HEAD request for docs links, with two retries (specified).
- Preserve the staged internal skill move and both symlinks (specified).
- Verify command sources against main at `40f17a958191ca81e05dc2c145d3d02712df4b7b`, fetched on September 8, 2026.
- No applicable design document exists. The swarm-apps design document describes a different system.

## Phase 1: Verify sources and write operator documentation

- [x] Resolve integration, embedding, and agent-fs minimums from code.
- [x] Replace the public skill placeholder and add detailed references.
- [x] Rewrite README and correct the Helm TL;DR.

### Verification

- Compare command source files with `git diff HEAD origin/main -- <source-files>`.
- Render the chart with `helm template swarm charts/agent-swarm --set auth.existingSecret=agent-swarm-secrets`.
- Validate Compose with `docker compose -f docker-compose.example.yml --env-file <scratch-env> config --quiet`.

## Phase 2: Guard distribution and document delivery

- [x] Add the public skill guard and wire it into package.json and Merge Gate.
- [x] Document the fourth delivery path in CLAUDE.md and runbooks/skills.md.
- [x] Preserve and verify the internal skill symlinks.

### Verification

- `bun run check:operator-skill`
- From a scratch directory: `npx -y skills@latest add /Users/taras/.t3/worktrees/agent-swarm/t3code-1acfa297 --agent claude-code -y`
- Confirm exactly one installed skill and its bundled reference files.

## Phase 3: Review and submit

- [x] Run required acceptance checks and applicable PR checks.
- [x] Review the diff with separate Standards and Spec agents. Resolve important findings.
Handoff: commit the verified changes, execute plugin/commands/create-pr.md, and stop for Taras's review.

### Verification

- `bun run lint`
- `bun run tsc:check`
- `git diff --check`
- Run the applicable commands from runbooks/ci.md before PR creation.

## Evidence

- Main command sources match this checkout. The published unversioned Helm command resolves chart `1.142.0`.
- `helm template` and Compose `config --quiet` pass with fixture credentials and generated UUIDs.
- README has 100 lines. Its Mermaid diagram matches the original exactly.
- Scratch installer found and installed exactly one skill, `agent-swarm`. All six Markdown files matched the source at installation.
- Scratch directory: `/private/tmp/agent-swarm-skill-install.X9EreJ`.
- A fresh Claude session invoked the Skill tool for `swarm-local-e2e` and loaded the original heading through the symlink.
- Root lint and typecheck pass. UI lint, build-mode typecheck, and E2E typechecks pass.
- The API contract suite passes: 9 scenarios. The UI suite passes: 40 tests, 17 expected skips.
- API, worker-slim, and eval Docker builds pass.
- Bun version, database boundary, API-key boundary, RBAC boundary, test-spawn boundary, audit columns, RBAC coverage, response coverage, and dependency graph checks pass.
- The live guard passed all 24 documentation URLs. Isolated fixtures cover retries, failure exit, fragment deduplication, malformed frontmatter, missing targets, and missing skill directory.
- The initial sandboxed test run could not bind local ports. A rerun with socket access exposed three tests affected by ambient provider credentials.
- Those three tests pass without ambient provider credentials. The complete isolated suite then reported 8342 passing, 12 skipped, and one Slack timestamp-ordering failure.
- The failing Slack file passes alone. A second complete run reproduced the same existing failure. At that point, no runtime or test code changed.

## Review decisions

- Standards: added the guard to the CI runbook and removed a single-use helper.
- Standards: retained the specified guard contract. Exactly-one installation is the acceptance check for this PR, not a ban on future public skills.
- Standards: retained the required network check in Merge Gate. Any tracked target can disappear in an unrelated file change.
- Standards minor: compact HTML keeps the README at 100 lines while preserving the original diagram and badge content.

- Spec review: no Critical, Important, or Minor findings.
- Final guard: 6 files, 31 tracked GitHub targets, 24 documentation URLs.
- Taras requested the Slack test fix before submission. The test now asserts exactly one new task by ID without relying on timestamp ordering.

## Slack test correction

- [x] Replace timestamp ordering with task identity and assert exactly one new task.
- [x] Verify the focused file: 25 passing tests.
- [x] Verify the complete unit suite after the correction: 8343 passed, 12 skipped, 0 failed.
- Standards and Spec reviews both approved the test correction. Runtime code and test inventory remain unchanged.

The final full run passed without concurrent Docker builds. A prior concurrent run timed out in three provider tests.
Final Docker builds, lint, and typecheck also pass. Final installer evidence: `/private/tmp/agent-swarm-skill-final.Ovpq1o`.

## Manual E2E

- `bun run e2e --only health,auth` checks the real API using the LOCAL_TESTING.md recipe.
- `curl -fsS http://localhost:3013/health` checks an operator deployment after Compose or Helm installation.
- `curl -fsS -H "Authorization: Bearer $API_KEY" http://localhost:3013/api/agents` checks agent registration.
- Fresh Claude session: confirm `/swarm-local-e2e` still resolves through the symlink without executing the deployment procedure.

Files in this skill

  • 2026-08-06-host-agnostic-mcp-registry.md2.1 KB
  • 2026-08-18-identity-budget-rejection-visibility.md1.9 KB
  • 2026-08-20-bun-14-adoption-pr.md7.4 KB
  • 2026-08-20-slack-channel-lifecycle.md2 KB
  • 2026-09-02-compose-pull-policy.md1.5 KB
  • 2026-09-04-agent-fs-upload-404-prepush.md2.5 KB
  • 2026-09-04-des-781-agent-browser-worker.md6.2 KB
  • 2026-09-07-provider-process-groups.md1.8 KB
  • 2026-09-08-acp-diagnostics.md2 KB
  • 2026-09-08-pages-authored-styles.md2.9 KB
  • 2026-09-08-public-operator-skill.md5.7 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.