All authors
Claude Skills by dancingteeth
github.com/dancingteeth29 skills0 installs1 views
- Batch Item Rubrics1. `loopBatchConfigSchema` parses mixed `loops` string and `{path,rubric}` entries. 2. Batch runner passes `batchRubric` into `runAgentLoop` when rubric is set. 3. `buildAgentLoopPrompt` renders `## Batch rubric` in the volatile tail when provided. 4. Existing string-only batches still work. 5. `tsc --noEmit` clean; focused vitest green.Votes: 0GitHub stars: 3
- Cli Max CostVerify cli-max-cost (maxCostUsd stop + HITL waiting).Votes: 0GitHub stars: 3
- Cli Resume ReportVerify cli-resume-report (resume hints, report links, Telegram delta).Votes: 0GitHub stars: 3
- Cli Setup DetectVerify cli-setup-detect (runtime probe + dump-before-write).Votes: 0GitHub stars: 3
- Cli WatchVerify the cli-watch live-progress loop. Use when implementing under .cursor/loops/cli-watch/.Votes: 0GitHub stars: 3
- Dsh PluginVerify the DeepSeek Harness companion plugin for Agent Looper.Votes: 0GitHub stars: 3
- Dsh SmokeRun: ```bash bash .cursor/loops/dsh-smoke/verify.sh ``` If it fails, create or fix `.cursor/loops/dsh-smoke/probe.txt` per `GOAL.md`, then rerun until exit 0.Votes: 0GitHub stars: 3
- Example FixRun the measurable verification checklist for an Agent Looper task. Use when implementing or fixing work under .cursor/loops/ and before declaring the loop complete. Never hand back partially verified work.Votes: 0GitHub stars: 3
- Loop Risk ProfilesMeasurable checks for configurable loopRisk profiles (TW `de4144f2-…`). 1. `pnpm exec tsc --noEmit` 2. `pnpm exec vitest run src/loop/loopRisk*.test.ts src/review/reviewPrompt.test.ts` 3. `pnpm exec vitest run` (full suite) 4. `agent-loop-review-preview` prints `Risk:` for example-fix loop On any failure: fix → rerun verify.sh. No partial handoff.Votes: 0GitHub stars: 3
- Meta ReviewVerify M5 cross-loop meta-review CLI for Agent Looper dogfood.Votes: 0GitHub stars: 3
- Opencode SmokeRun the shell verifier: ```bash bash .cursor/loops/opencode-smoke/verify.sh ``` If it fails, create or fix `.cursor/loops/opencode-smoke/probe.txt` per `GOAL.md`, then rerun until exit 0.Votes: 0GitHub stars: 3
- Pi SmokeRun: ```bash bash .cursor/loops/pi-smoke/verify.sh ``` If it fails, create or fix `.cursor/loops/pi-smoke/probe.txt` per `GOAL.md`, then rerun until exit 0.Votes: 0GitHub stars: 3
- Post Success ReviewVerify extract post-success review refactor for Agent Looper.Votes: 0GitHub stars: 3
- Reproduce AgentVerify M2b fresh-context reproduce agent for Agent Looper dogfood.Votes: 0GitHub stars: 3
- Reproduce Before ReportVerify M2 reproduce-before-report phase 2a for Agent Looper dogfood.Votes: 0GitHub stars: 3
- Secondary JudgeVerify M3 multi-family secondary review judge for Agent Looper dogfood.Votes: 0GitHub stars: 3
- Verify ModeVerify M4 Track B verifyMode harness for Agent Looper dogfood.Votes: 0GitHub stars: 3
- DocsTrack **A** of roadmap milestone **M4** ([`loop-review-roadmap.md`](./loop-review-roadmap.md)). Taskwarrior: `fe3f4076-b997-4d28-a59a-baf720c28e5d` (project `agent-loop`).Votes: 0GitHub stars: 3
- Check Running LoopsChecks whether agent-loop / agent-loop-batch jobs are actually alive, stale, hung, dead, or done, for any worker/judge runtime (Cursor, Cline, OpenCode, Pi, Codex, DSH, Muse, Claude). Use when the user asks if loops are ok, stuck, running, healthy, or what the current batch status is; when they mention agent-loop, loop-batch, verify.sh harness, or AGENT_LOOP_DONE.Votes: 0GitHub stars: 3
- Design LoopDesign a measurable Agent Looper cycle — frozen GOAL.md, verify.sh finish line, unknowns preflight, and scope limits. Use when creating or tightening a loop before running Agent Looper, or when a finished loop was still wrong and verify needs revision from export packs.Votes: 0GitHub stars: 3
- Install Agent LooperInstall and wire @dancingteeth/agent-looper in a consumer repo — dependency, agent-loop-init, scripts, and first dogfood run. Use when adding the harness or debugging why the CLI is missing.Votes: 0GitHub stars: 3
- Review GateConfigure Agent Looper post-verify quality review and reviewGate — Proceed / Guide / Deny / Confirm, impact-severity gating, and when not to thrash another iteration. Use when writing REVIEWS.md or tuning postQualityReview.Votes: 0GitHub stars: 3
- Design LoopFreeze GOAL.md + verify.sh + loop.json only. Do not implement, SSH-walk production, or dump secrets. Use when the user wants an Agent Looper for a task, or when a finished loop was still wrong and verify needs revision from export packs (still freeze-only).Votes: 0GitHub stars: 3
- Install Agent LooperInstall the Agent Looper CLI in the current repo. Use only when `agent-loop --help` is missing. Do not dump Doppler or OpenCode secrets.Votes: 0GitHub stars: 3
- Review GateConfigure Agent Looper reviewGate / REVIEWS.md. Use only when writing the judge spec.Votes: 0GitHub stars: 3
- Run Loop In DshAfter a loop is frozen, start agent-loop as DSH background bash. Use when the user asks to grind from dsh web. Do not foreground-bash the grind.Votes: 0GitHub stars: 3
- Run Loop In OpencodeInstall Agent Looper and start a frozen loop from OpenCode with the agent_loop_start tool. Use when the user asks to add the harness, grind, run, or implement a loop from an OpenCode session. Do not foreground-bash the grind.Votes: 0GitHub stars: 3
- TemplatesRun the measurable verification checklist for an Agent Looper task. Use when implementing or fixing work under .cursor/loops/ and before declaring the loop complete. Never hand back partially verified work.Votes: 0GitHub stars: 3
- Hello VerifyTiny example skill discovered via Agent Plugins skills/ layout.Votes: 0GitHub stars: 3