Utility commands

Utility commands handle recovery, cleanup, and lifecycle operations. They work independently of the main pipeline.

Command	Type	What it does
/agents fix	LLM	Retries a failed task using failure context from prior attempts
/agents revert	Bash	Undoes tactical planning, restores the feature issue to its pre-plan state
/agents close	Bash	Full cleanup: closes issues, PRs, and deletes branches

---

/agents fix

flowchart LR
  Triggers@{ shape: bolt, label: "Triggers" }
  FixAgent["Fix agent"]

  Triggers --triggers--> FixAgent

  class Triggers triggersOrange
  class FixAgent fixRed

  classDef triggersOrange fill:#ffe8d4,stroke:#d66a28,color:#d66a28
  classDef fixRed fill:#ffd4d4,stroke:#d62828,color:#d62828

The Fix agent recovers from failed execution attempts. It reads the task spec, inspects the failure context from recent comments, and resumes or replaces the prior implementation.

Script: .autoducks/agents/fix/pre.sh + post.sh
Trigger: Issue comment /agents fix on any task issue

Behavior

1. Reacts to the trigger comment with 👀.
2. Searches for an existing partial branch from a prior failed attempt.
   • If found: checks out that branch to continue where the previous attempt left off.
   • If not found: creates a new fix branch feature/{feature_num}-issue-{issue_num}-fix-{timestamp}.
3. Prepares the task spec and failure context from the last 10 comments.
4. The LLM reads both, then fixes or completes the implementation.
5. Commits and pushes. Reuses an existing PR if present; otherwise creates Fix: {title}.
6. If feature-parented: auto-merges and triggers the wave orchestrator.
7. Reacts with 👍 and confirms the fix.

Tip

Before running /agents fix, add a comment to the task issue explaining what went wrong. The agent reads the last 10 comments as context, so more detail leads to better results.

---

/agents revert

flowchart LR
  Triggers@{ shape: bolt, label: "Triggers" }
  RevertAgent["Revert agent"]

  Triggers --triggers--> RevertAgent

  class Triggers triggersOrange
  class RevertAgent revertYellow

  classDef triggersOrange fill:#ffe8d4,stroke:#d66a28,color:#d66a28
  classDef revertYellow fill:#fff3cd,stroke:#856d00,color:#856d00

Undoes the work of the Tactical agent: closes all task issues and restores the feature issue body to how it looked before /agents devise was run.

Script: .autoducks/agents/revert/run.sh
Type: Bash — no LLM
Trigger: Issue comment /agents revert on a Feature issue

Behavior

1. Reacts with 👀.
2. Extracts task numbers from the YAML waves: block in the feature issue body.
3. Closes all task issues with a comment referencing the revert.
4. Removes Ready and draft labels from the feature issue.
5. Restores the original issue body from edit history (last non-bot edit).
6. Deletes all bot comments on the feature issue.
7. Reacts with 👍.

Does not delete branches or close PRs — use /agents close for that.

Use this when the tactical plan needs to be completely redone. After reverting, run /agents devise again.

---

/agents close

flowchart LR
  Triggers@{ shape: bolt, label: "Triggers" }
  CloseAgent["Close agent"]

  Triggers --triggers--> CloseAgent

  class Triggers triggersOrange
  class CloseAgent closeGray

  classDef triggersOrange fill:#ffe8d4,stroke:#d66a28,color:#d66a28
  classDef closeGray fill:#e8e8e8,stroke:#586069,color:#586069

Full teardown of a feature and all associated work.

Script: .autoducks/agents/close/run.sh
Type: Bash — no LLM
Trigger: Issue comment /agents close on a Feature issue

Behavior

1. Reacts with 👀.
2. Extracts task numbers from the YAML waves: block.
3. For each task: finds branches, closes open PRs, deletes branches, closes the task issue.
4. Closes the feature PR (if open) and deletes the feature branch.
5. Closes the feature issue with a cleanup summary.
6. Reacts with 👍.

What gets removed

Object	Action
Task issues	Closed
Task PRs	Closed
Task branches	Deleted
Feature branch	Deleted
Feature PR	Closed
Feature issue	Closed

Caution

/agents close is destructive. Deleted branches cannot be easily recovered. Use /agents revert if you only want to undo the plan while keeping the feature issue open.