addy-auto-unblock
Guides safe autonomous recovery while /addy-auto mode is active.
Skill by teknologist
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/teknologist-addy-auto-unblock
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/addy-auto-unblock
What's inside
- Skill name
addy-auto-unblock- Description
- Guides safe autonomous recovery while /addy-auto mode is active. Use only when Addy Auto Mode is active and a build, verify, review, fix, commit, or plan-continuation step is blocked or repeats without completing.
- License
- MIT
- Instructions
- 59 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Addy Auto Unblock Use this only inside `/addy-auto`. ## Non-negotiable rule Do not trade correctness for progress. Never skip, weaken, or silently reinterpret acceptance criteria, verification, or review just to unblock the loop. `[x] Verified` and `[x] Reviewed` require real evidence from this run. ## Recovery loop Before pausing: 1. Load and apply `debugging-and-error-recovery`. 2. Reproduce the blocker with the exact failing command, review finding, missing artifact, or git state. 3. Classify the blocker: - implementation defect - test/fixture/tooling gap - build-time missing ADR/spec/steering context that can be repaired by linking an existing unambiguous artifact - review finding needing a code or test fix - ADR-related review finding needing implementation repair, missing spec/plan required context, or an existing ADR link - plan/status checkbox out of sync with evidence - workflow state/stat synchronization lag where the plan already has the owned checkbox checked and this run has real phase evidence - genuinely unsafe or ambiguous decision 4. Fix the root cause when the fix is safe and scoped to the current task. 5. Add or update meaningful regression coverage when behavior changed. 6. Re-run the required verification/review step. 7. Update only the lifecycle checkbox owned by the completed step. ADR-related review findings are recoverable when the safe scoped fix is to preserve an existing ADR, restore a skipped `must not` guardrail, or add missing references to an already-existing ADR in the spec/plan required context. Pause instead of auto-fixing when resolution requires creating or superseding an ADR, choosing between conflicting ADRs, or making a product/security/architecture judgment. Build-time missing ADR/spec/steering context is recoverable only when the existing ADR or steering file is unambiguous and the safe scoped fix is limited to linking it from the active spec/plan required context. After repairing that context, rerun the blocked build step. Pause instead of redesigning during build or guessing; if re-defining the spec, a new/superseding ADR, conflicting ADR interpretation, or human architecture/product/security judgment is needed, stop and report the blocker. If the repeated blocker is "missing lifecycle evidence: Reviewed" but the active task already has `[x] Reviewed` and the latest real `/addy-review` for that task reported `No issues found`, do not re-run review just to satisfy stale state. Treat it as synchronization lag, preserve the review evidence in the report, and let Addy auto commit or advance to the next unfinished slice. ## Missing or failing verification artifacts Missing test files, fixtures, commands, snapshots, generated docs, or local setup are not automatic blockers. - If the artifact is required by the current acceptance criteria, create or repair it with meaningful assertions. - If the artifact is stale or over-broad, preserve the acceptance criteria and replace it only with equivalent or stronger verification evidence. Record why in the plan. - Do not create placeholder tests, empty fixtures, or assertions that cannot fail when the required behavior regresses. ## Auto-dispatched fix-all handoff When `/addy-fix-all` runs inside `/addy-auto`, it is only the fix pass. Fix the surfaced review issues and run narrow validation for the changed scope, then stop. Do not invoke or perform `/addy-verify` or `/addy-review` inside that fix-all turn. The auto monitor owns the follow-up order: `/addy-verify`, then `/addy-review`. ## When to pause Pause only after the recovery loop when: - the next step is destructive, external, or needs credentials the agent does not have; - the acceptance criteria conflict or require product/security/architecture judgment; - the only apparent fix would weaken verification or review; - the failure cannot be reproduced with available evidence; - repeated root-cause fixes fail and further attempts would be speculative. When pausing, report the exact blocker, evidence gathered, commands run, and the safest next action.
From teknologist/pi-addy-workflow, MIT.
More from teknologist/pi-addy-workflow
- 0.469score
debugging-and-error-recovery
teknologist
Diagnose failures systematically.
SkillCFreshness: CClaude Code +9 - 0.469score
using-addy-workflow
teknologist
Governs the Addy workflow package lifecycle.
SkillCFreshness: CClaude Code +9 - 0.469score
spec-driven-development
teknologist
Create a concise spec before implementation.
SkillCFreshness: CClaude Code +9 - 0.469score
planning-and-task-breakdown
teknologist
Break specs into ordered, verifiable tasks.
SkillCFreshness: CClaude Code +9 - 0.464score
incremental-implementation
teknologist
Implement plans in small safe slices.
SkillCFreshness: CClaude Code +9 - 0.464score
performance-optimization
teknologist
Review and improve latency, throughput, resource use, rendering, and scalability.
SkillCFreshness: CClaude Code +9 - 0.464score
security-and-hardening
teknologist
Review untrusted input, auth, secrets, data exposure, dependencies, and deployment risks.
SkillCFreshness: CClaude Code +9 - 0.464score
shipping-and-launch
teknologist
Prepare go/no-go release decisions with verification and rollback.
SkillCFreshness: CClaude Code +9 - 0.464score
systematic-debugging
teknologist
Reproduce, minimize, hypothesize, instrument, fix, and regression-test bugs.
SkillCFreshness: CClaude Code +9 - 0.464score
browser-testing-with-devtools
teknologist
Verify browser behavior with real runtime data when UI or browser code is involved.
SkillCFreshness: CClaude Code +9 - 0.464score
code-review-and-quality
teknologist
Review changes for correctness, readability, architecture, security, and performance.
SkillCFreshness: CClaude Code +9 - 0.464score
code-simplification
teknologist
Simplify code without behavior changes.
SkillCFreshness: CClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- May 27, 2026 (5 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/teknologist/pi-addy-workflow
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from teknologist
- 0.473score
crit-cli
teknologist
Use when replying to Crit comments, inspecting Crit status, or interpreting Crit review JSON from Pi.
SkillCFreshness: CClaude Code +9 - 0.469score
debugging-and-error-recovery
teknologist
Diagnose failures systematically.
SkillCFreshness: CClaude Code +9 - 0.469score
planning-and-task-breakdown
teknologist
Break specs into ordered, verifiable tasks.
SkillCFreshness: CClaude Code +9 - 0.469score
spec-driven-development
teknologist
Create a concise spec before implementation.
SkillCFreshness: CClaude Code +9 - 0.469score
using-addy-workflow
teknologist
Governs the Addy workflow package lifecycle.
SkillCFreshness: CClaude Code +9 - 0.469score
crit
teknologist
Use when Crit review comments are injected into Pi context or when the user asks to start a Crit review from Pi.
SkillCFreshness: CClaude Code +9
Cite this page
addy-auto-unblock skill. Agent Mart. https://agentmart.dev/skills/teknologist-addy-auto-unblock (last updated 2026-10-24)