brdohman/error-recovery
.claude/skills/workflow/error-recovery/SKILL.md
Structured recovery templates for common agent failure modes. Use when an agent fails mid-task due to context exhaustion, build failure, or unclear requirements.
$ install --global
skillsauthnpx skillsauth add brdohman/agile-maestro error-recoveryInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 16, 2026, 7:20 PM8.8s1 file scanned
SKILL.md
- name:
- error-recovery
- description:
- Structured recovery templates for common agent failure modes. Use when an agent fails mid-task due to context exhaustion, build failure, or unclear requirements.
- user-invocable:
- false
Error Recovery Templates
When an agent fails, the coordinator spawns a NEW agent with a recovery prompt. The recovery prompt must be under 500 words and include only what the new agent needs.
Context Exhaustion
Agent ran out of context before completing the task.
Recovery prompt template:
Task ID: [task-id]
Story: [story-id]
What was attempted: [1-2 sentences from the last implementation comment, or STARTING comment if no impl comment exists]
What remains: [list unchecked items from task checklist]
Files already created/modified: [list from git diff or last comment]
Error: Agent exhausted context before completing.
Pick up where the previous agent left off. Read the task via TaskGet, check what files exist, and complete the remaining checklist items. Do NOT redo work that's already done.
Key rule: Never paste the failed agent's full output. Summarize in 2-3 sentences.
Build Failure
Agent's code doesn't compile.
Recovery prompt template:
Task ID: [task-id]
Build error: [paste ONLY the error message, not full build log — typically 3-10 lines]
File with error: [path]
What the task is doing: [1 sentence from task description]
Fix the build error. Read the file, fix the issue, rebuild. If the fix requires understanding other files, read only what's needed.
Key rule: Paste the error message, not the full xcodebuild output. The error is usually 3-10 lines. The full log can be 500+ lines.
Test Failure
Tests fail after implementation.
Recovery prompt template:
Task ID: [task-id]
Failing test: [TestClassName/testMethodName]
Test error: [assertion failure message — 1-3 lines]
Test file: [path]
Implementation file: [path]
Read the failing test, understand what it expects, read the implementation, and fix the mismatch. Run the test again to verify.
Unclear Requirements
Agent couldn't determine what to build from the task description.
Recovery approach: Don't spawn a recovery agent. Instead:
- Read the task description
- Identify what's ambiguous
- Add a comment to the task:
type: "note", content: "BLOCKED: [what's unclear]" - Set
metadata.blocked: true - Ask the user to clarify via AskUserQuestion
Merge Conflict
Agent's changes conflict with another agent's concurrent work.
Recovery prompt template:
Task ID: [task-id]
Conflicting files: [list from git status]
This task's intent: [1 sentence]
Resolve the merge conflicts in the listed files. Keep both sides' changes where possible. If the conflict is semantic (both sides changed the same logic), read both versions and pick the correct one based on the task description. Rebuild after resolving.
Recovery Rules
- Max 500 words in the recovery prompt. Anything longer defeats the purpose.
- Never paste full error logs. Extract the relevant 3-10 lines.
- Never paste file contents. The recovery agent can read files itself.
- Include the task ID so the recovery agent can TaskGet for full context.
- State what's done and what remains. Don't make the recovery agent figure out where the previous agent stopped.
Related Skills
brdohman/xctest-patterns
testing
VerifiedTrustedCommunity
XCTest patterns for macOS Swift apps. Unit tests, async tests, Core Data tests, mock patterns, and assertion reference. Use when writing or reviewing tests.
2SKILL.mdUpdated Apr 16, 2026
brdohman/workflow-state
tools
VerifiedTrustedCommunity
How to transition workflow state between review stages. Rules for setting review_stage and review_result fields on Stories and Epics.
2SKILL.mdUpdated Apr 16, 2026
brdohman/workflow-comments
documentation
VerifiedTrustedCommunity
Comment structure and rules for task workflow updates. Use when adding any comment to a task during implementation, review, or fix cycles.
2SKILL.mdUpdated Apr 16, 2026
brdohman/validate-task
testing
VerifiedTrustedCommunity
Validate task/story/epic/bug/techdebt metadata against schema v2.0. Run after TaskCreate or TaskUpdate to verify compliance. Returns pass/fail with actionable details.
2SKILL.mdUpdated Apr 16, 2026