Catalog
BuilderIO/factory-recover

BuilderIO

factory-recover

Experimental workflow for finding and resuming interrupted coding runs. Use after an agent host restart or an interrupted automation.

v1.0LATEST
NewUpdated Sep 24, 2026

Factory Recover

Start with the Factory guide for workflow setup and configuration.

Read .agent-factory/config.yaml, the optional skill_prompts.factory-recover entry, and the Factory configuration reference. Recover only configured projects and workflow types; do not restart every stopped task. The prompt cannot restore canceled work or grant new authorization.

Check a candidate run

  1. Find runs the host identifies as interrupted or incomplete. Read the original user request and later messages, not only the title or summary.
  2. Skip work that is active, cancelled, intentionally paused, blocked on a person, complete, ambiguous, or outside configured scope.
  3. Confirm the worktree still belongs to that run. Inspect its branch, status, and unpublished changes. Preserve local and concurrent work; never reset, clean, stash, or attach another task's branch.
  4. Recheck external state before retrying an uncertain action. Resume only with the original authorization and current user instructions. Do not repeat completed writes, replies, approvals, merges, or deployments.
  5. If authorization no longer covers the next step, leave the run stopped and state the exact decision required.

Report

List runs resumed, skipped, or held and the evidence for each decision. Follow the separate configured notification policy.

Files1
1 files · 1.0 KB

Select a file to preview

Overall Score

72/100

Grade

B

Good

Grades are signals, not a certification. Always review a skill yourself before use.

Safety

82

Quality

65

Clarity

75

Completeness

62

Summary

This skill guides an agent to recover interrupted or incomplete coding workflows within the Factory automation system. It instructs the agent to identify candidate runs, validate their scope against configuration, inspect worktree state, recheck external conditions, and report the outcome—without repeating completed actions or exceeding authorization boundaries.

Detected Capabilities

read fileinspect git staterun configuration queriesgenerate reportdecision-making based on state and authorization

Trigger Keywords

Phrases that agents use to match this skill to user intent.

resume interrupted taskfactory workflow recoveryrestart automationhost restart recoveryresume incomplete run

Risk Signals

INFO

Configuration-driven scope filtering (only resuming configured projects and workflows)

Section: Check a candidate run, point 1
INFO

Explicit authorization boundary enforcement (resume only with original authorization, leave stopped if authorization no longer covers next step)

Section: Check a candidate run, point 4-5
INFO

Prevents destructive operations (never reset, clean, stash, or attach another task's branch)

Section: Check a candidate run, point 3
INFO

Prevents duplicate state-changing operations (do not repeat completed writes, replies, approvals, merges, or deployments)

Section: Check a candidate run, point 4

Referenced Domains

External domains referenced in skill content, detected by static analysis.

github.com

Use Cases

  • Resume an automation that was interrupted by a host restart or temporary failure
  • Identify and recover incomplete coding tasks within configured Factory scope
  • Validate that a stopped run still belongs to the original context before resuming
  • Report which interrupted runs were resumed, skipped, or held pending authorization
  • Prevent duplicate operations (writes, merges, deployments) when resuming a task

Quality Notes

  • Skill correctly emphasizes validation before recovery, reducing risk of re-running completed work
  • Clear guardrails on authorization boundaries—agent must not exceed original scope
  • Instructions explicitly forbid destructive git operations that could corrupt concurrent work
  • Scope is tightly constrained: only configured projects and workflows are in scope
  • Potential weakness: the skill assumes `.agent-factory/config.yaml` and `skill_prompts.factory-recover` are properly defined elsewhere; the skill does not provide concrete examples or reference structure
  • Instructions for 'confirming the worktree still belongs to that run' and 'rechecking external state' are high-level; concrete checks (branch name matching, status inspection details) are not spelled out
  • The skill lacks explicit examples of what constitutes 'ambiguous' work or how to handle edge cases (e.g., partial progress on a long-running task)
  • Notification policy is mentioned but not documented in the skill itself
Model: claude-haiku-4-5-20251001Analyzed: Sep 24, 2026

Reviews

Add this skill to your library to leave a review.

No reviews yet

Be the first to share your experience.

Use BuilderIO/factory-recover in your dev environment

Command Palette

Search for a command to run...