// Run this skill when a Flow Healer run fails or behaves unexpectedly and the user wants a fast, deterministic diagnosis. Use for requests like "triage this healer failure", "why did the run fail", "classify this issue state", or "tell me if this is an environment problem or a product bug".
Review a local codebase, turn validated findings into small GitHub issues, and order them by fix path. Use when the user wants a code review that produces GitHub issues instead of code changes, especially for requests like "scan this repo and create issues", "do a multi-agent review", "open fix-path issues", or "review without editing code".
Activates a structured multi-agent workflow (Team Lead + specialized Employees) to handle complex tasks that require decomposition, specialized execution, and synthesis. Use when a user request: (1) Involves multiple distinct domains (e.g., research, coding, testing), (2) Is large enough to be split into parallel sub-tasks, (3) Requires high-quality synthesis of diverse inputs. Triggers on requests like: 'Research 3 competitors...', 'Build a feature with backend/frontend/testing...', 'Analyze a business idea with market/risk/monetization agents...'
Run this skill when Flow Healer triage points to connector or patch generation failures and the user wants a deterministic connector-focused debug pass. Use for requests like "debug the connector", "why did the proposer emit no patch", "validate the verifier payload", or "compare proposer and verifier output contracts".
Run this skill when the user wants a guarded live Flow Healer smoke test on a real GitHub repo. Use for requests like "do a live smoke", "test this repo live", "open a real PR through the healer", or "validate issue-to-PR plumbing on GitHub".
Run this skill when the user wants safe local validation before any live Flow Healer action. Use for requests like "run local checks", "validate the healer locally", "do a dry run first", or "make sure the repo is healthy before live GitHub actions".
Run this skill when the user wants to validate or execute a Flow Healer PR follow-up on an existing live PR. Use for requests like "rerun the same PR", "ingest review feedback", "follow up on this healer PR", or "reuse the existing issue and PR after comments".
| name | flow-healer-triage |
| description | Run this skill when a Flow Healer run fails or behaves unexpectedly and the user wants a fast, deterministic diagnosis. Use for requests like "triage this healer failure", "why did the run fail", "classify this issue state", or "tell me if this is an environment problem or a product bug". |
Use this skill after any failed or suspicious Flow Healer attempt.
--db-path--issue-idThe script emits a JSON object with:
issuelatest_attemptdiagnosisdiagnosislatest_attempt.failure_classlatest_attempt.failure_reasonissue.stateconnector_or_patch_generation as an immediate handoff to flow-healer-connector-debug.scripts/triage_issue.py with the healer DB path and issue id.diagnosis first, then confirm it against latest_attempt.failure_class, latest_attempt.failure_reason, and issue.state.product_bug, read references/failure_catalog.md and compare against known incidents.operator_or_environment, fix the environment before another live run..venv/bin/python skills/flow-healer-triage/scripts/triage_issue.py \
--db-path /tmp/flow-healer-state/repos/live/state.db \
--issue-id 3
operator_or_environmentrepo_fixture_or_setupconnector_or_patch_generationproduct_bugexternal_service_or_githuboperator_or_environment: repair the environment and rerun flow-healer-preflight.repo_fixture_or_setup: repair the repo/setup and rerun flow-healer-local-validation.connector_or_patch_generation: hand off to flow-healer-connector-debug and compare proposer versus verifier output contracts.product_bug: capture evidence and escalate.external_service_or_github: retry later with an operator note.product_bug or when the pattern is unfamiliar.