bmad-loop-resolve

Resolves CRITICAL escalations in paused bmad-loop runs by updating frozen specs with human decisions.

1|Updated Sep 18, 2026
One-click install
npx skills add https://github.com/PastaSus/egg-defender --skill bmad-loop-resolve-pastasus
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: bmad-loop-resolve
Source: https://github.com/PastaSus/egg-defender/tree/main/.agents/skills/bmad-loop-resolve
Command: npx skills add https://github.com/PastaSus/egg-defender --skill bmad-loop-resolve-pastasus

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When an automated bmad-loop run pauses because a dev or review session hit a CRITICAL escalation — a contradiction or gap in the frozen spec it could not safely resolve alone — this Skill provides an interactive workflow to disambiguate the spec with a human and unblock the run. ## Core Features & Use Cases - Escalation Context Loading: Reads the orchestrator-written context.json to understand the paused run, the escalated story, and the exact ambiguity reported. - Interactive Spec Resolution: Presents the contradiction to the human with 2-4 concrete resolution options, then amends the frozen spec's <frozen-after-approval> block so a fresh dev session has exactly one correct reading. - Patch Restore Handling: For review-stage intent gaps with a saved patch, supports resuming review on the already-written change via a restore_patch marker instead of re-implementing from scratch. - Sentinel & Ambiguous-File Resolution: Handles stories-mode sentinels (unplannable stories) and duplicate story-file matches by fixing upstream SPEC.md or stories.yaml entries. - Use Case: A review session escalates because the frozen spec never said whether a CLI list command should output JSON or plain text. You invoke this Skill, review the escalation with the human, amend the spec to mandate JSON, and write the resolution marker so the orchestrator re-arms and re-drives the story. ## Quick Start Invoke the skill with /bmad-loop-resolve <story-key> after a bmad-loop run pauses on a CRITICAL escalation, then work with the human to resolve the ambiguity.

Frequently Asked Questions about bmad-loop-resolve

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I resolve a CRITICAL escalation in a paused bmad-loop run?

Invoke /bmad-loop-resolve with the escalated story key in an interactive session. Read the context.json at $BMAD_LOOP_RESOLVE_CONTEXT, present the ambiguity to the human with resolution options, amend the frozen spec, and write the resolution marker at resolution_path.

What is a frozen spec and why do dev sessions escalate on it?

A frozen spec is the authoritative story specification in its <frozen-after-approval> block that dev and review sessions treat as final. Sessions escalate when that block is silent on, or contradicts, a case the implementation actually hit, making safe autonomous work impossible.

When should I use restore_patch in the resolution marker?

Use restore_patch only when a review-stage intent gap saved a patch and the human confirms the attempted implementation reading was correct. The orchestrator then re-applies the patch and resumes review. Omit it when restore_supported is false or the attempted reading was wrong.

Can this skill implement the story or change the spec status field?

No. The skill only edits spec content and writes the resolution marker. It must not write result.json, change sprint-status.yaml, set the spec status field, implement code, run tests, or commit — the orchestrator handles re-arming deterministically.

What happens if the escalation cannot be resolved in the session?

If the human defers or the fix is out of scope, do not write the resolution marker. The story stays escalated and the run remains paused, which is the safe default — the orchestrator never re-arms a story without a recorded resolution.