| name | hipfire-autoheal |
| description | Triage and repair hipfire runtime failures such as daemon hangs, stale serve.pid, port 11435 conflicts, ROCm include-path problems, missing precompiled kernels, VRAM OOM, kernel JIT failures, and multi-turn recall regressions. Use after diagnostics identify a likely runtime issue or when the user asks to fix a broken hipfire serve/run flow. |
hipfire-autoheal
Use this skill after hipfire-diag or when the symptom is clearly a runtime
failure. Start with evidence, apply the smallest targeted fix, then verify the
specific user-facing path that failed.
Workflow
- Gather triage:
.agents/skills/hipfire-autoheal/triage.sh
.agents/skills/hipfire-autoheal/triage.sh --json
- Read
playbook.md and work the fix catalog in order unless the evidence
clearly points to a later item.
- Check
known-issues.md for hardware/model-specific caveats.
- Use
bisection.md only for repeatable hangs or regressions that survive
the standard fix catalog.
Guardrails
- Ask for approval before killing user processes, deleting files, changing
persistent config, installing packages, rebooting, or running privileged
commands.
- Prefer one-run environment overrides while bisecting, for example
HIPFIRE_KV_MODE=q8 or HIPFIRE_ATTN_FLASH=never, before changing config.
- If the fix changes code or runtime behavior touching kernels, quant formats,
dispatch, fusion, rotation, rmsnorm, or spec-decode, run
./scripts/coherence-gate-dflash.sh before claiming the repair is correct.
