om-fix-issue — Claude Code skill: what it does, how to install
om-fix-issue — a skill for Claude Code and Codex that diagnoses a tracker issue, verifies a regression test, and leaves staged changes for manual review without committing or opening a PR.
MIT 502 installs ★ 282 updated 2026-10-22
npx skills add open-mercato/skills --skill om-fix-issue -a claude-codeWhat om-fix-issue does
The skill orchestrates lower-level companions to resolve tracker bugs safely. It begins by running om-verify-in-repo inside an isolated worktree. If action is needed, it calls om-root-cause, requires a confirmed failing regression run, and freezes test paths into .om-freeze-tests so implementation loops cannot weaken test coverage.
Code changes proceed through om-fix and get inspected by om-code-review. With multi or high-assurance execution profiles, reviews execute across concurrent advisor models using fresh contexts. Crucially, the workflow never commits, pushes, or opens a pull request: it halts once changes are indexed in the git stage with validation artifacts ready for human inspection.
Execution halts if prerequisite companions (om-verify-in-repo, om-root-cause, om-fix, om-code-review) are missing. When triggered with --abort or upon fatal errors, the harness releases the tracker assignment and clears the in-progress label.
What it helps with
- verify tracker issues inside clean isolated worktrees
- enforce red-to-green verification with frozen regression test suites
- run multi-model independent code reviews in clean contexts
- stage verified issue fixes without pushing code or creating pull requests
Example prompts for the agent
- Fix this issue 142 but stop before PR
- Staged issue fix for issue PROJ-88
- Run om-fix-issue 405 --profile high-assurance
How to install om-fix-issue
1Skills CLI (Vercel)
npx skills add open-mercato/skills --skill om-fix-issue -a claude-code2Manually: put the files in place
~/.claude/skills/om-fix-issue/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/om-fix-issue/
1Skills CLI (Vercel)
npx skills add open-mercato/skills --skill om-fix-issue -a codex2Manually: put the files in place
.agents/skills/om-fix-issue/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add open-mercato/skills --skill om-fix-issue -a cursor2Manually: put the files in place
.cursor/skills/om-fix-issue/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install open-mercato/skills skills/om-fix-issue2Manually: put the files in place
.github/skills/om-fix-issue/SKILL.md1Skills CLI (Vercel)
npx skills add open-mercato/skills --skill om-fix-issue -a gemini-cli2Platform CLI
gemini skills install https://github.com/open-mercato/skills --path skills/om-fix-issueManually: put the files in place
.devin/skills/om-fix-issue/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add open-mercato/skills --skill om-fix-issue -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about om-fix-issue
How to install om-fix-issue in Claude Code?
Run npx skills add open-mercato/skills --skill om-fix-issue -a claude-code.
What does om-fix-issue do?
It diagnoses an issue, enforces a red regression test, generates the fix, runs automated review, and leaves verified files staged for human review without pushing.
What is required to run om-fix-issue?
Your setup needs the companion skills available: om-verify-in-repo, om-root-cause, om-fix, and om-code-review.
Source text
SKILL.md · MIT
---
name: om-fix-issue
description: Qualify and fix a tracker issue through the existing lower-level analysis, implementation, test, and review skills, then leave a verified staged branch for human review without committing, pushing, or opening a pull request. Use when the user says "fix this issue but stop before PR", "leave the fix staged", "staged issue fix", "napraw issue bez PR", or "zostaw zmiany staged".
---
# Fix Issue for Human Review
Drive a real issue fix to a staged, reviewable handoff. Use the configured
harness profile for optional workers and independent reviewers while leaving all
existing pipeline skills unchanged.
## Arguments
- `--interactive` (default) or `--autonomous` — apply the execution mode throughout this workflow; never enable publication. Reject both together.
- `{issueId}` (required) — tracker issue identifier.
- `{repo}` (optional) — repository handle understood by the tracker descriptor.
- `--profile <standard|optimized|multi|multi-optimized|high-assurance>` (optional) — default `standard`.
- `--force` (optional) — override an active claim only after explicit user confirmation.
- `--abort` (optional) — release this harness run's issue claim without publishing.
## Chaining
Consume the issue or brief and requested profile. Invoke the base workflow and its lower-level companions by name; stop with the missing name if one is unavailable. Emit the staged handoff, optional `Issue:`/`Spec:` references, and final `WORKTREE=` line. Publication is a separate human-authorized step.
## Workflow
**Claims:** `references/claim-pr.md` governs atomic/fallback ownership and release.
**ALWAYS check first:** Apply `.ai/skills/om-fix-issue/SKILL.md` when present; safety rules still win.
0. **Agentic setup** — follow `references/agentic-setup.md` before any action: load approved configuration, apply local extensions and shared rules, and preserve the untrusted-content boundary.
### 1. Preflight and trusted profile
Follow `references/issue-workflow.md` to enforce claim concurrency, resolve a
trusted base-revision harness config plus optional user-local overlay, validate
the requested profile, and probe its selected models.
### 2. Qualify the issue
Run `om-verify-in-repo` verbatim. Stop on `NO_ACTION_NEEDED`, citing its
evidence. Do not claim the issue before this gate says work is needed.
### 3. Isolate and diagnose
Create an isolated worktree and capture the harness start-state artifact. Run `om-root-cause`
verbatim. For `multi`, `multi-optimized`, and `high-assurance`, review the root-cause brief through the harness
before implementation using the bound `om-code-review` council contract: a
fresh Claude context plus every configured advisor receives the same immutable
diagnosis packet. Re-investigate any confirmed diagnosis blocker.
### 4. Implement and verify
Invoke `om-fix` with the root-cause brief before any implementation. It owns
the issue claim and observed failing regression first. Inside its implementation
phase, optimized profiles dispatch bounded worker packets; `high-assurance`
uses manifest packets with blind review, fresh verification, and exact-diff
gates. Resume that owning workflow for integration, validation, and self-review.
Require the regression test to be observed failing for
the diagnosed reason before the fix is applied, and record that red run in the
run artifacts. When the prevention hooks are installed, write the exact test
paths as a JSON array to `.om-freeze-tests` at the worktree root after the red run and remove it before
staging, so the fix loop cannot edit tests. A worker packet never replaces
that gate.
### 5. Review and preview
Run `om-code-review` on the complete uncommitted diff. For `standard` and
`optimized`, create a new Claude context with no inherited implementation
transcript and run the skill there. For `multi`, `multi-optimized`, and
`high-assurance`, follow the `om-harness` bound council contract: freeze the
exact diff, validation evidence, and repository rules; start fresh Claude and
every configured advisor concurrently with the same complete `om-code-review`
rubric and packet; then require Claude's matching artifact before
reconciliation. Reconcile every finding against code or
runtime evidence. Every configured council reviewer must complete — the runtime retries failures and a council missing any required reviewer yields no verdict; re-run it, repair the binding, or stop for the user's decision. For UI changes, run local verification through
`om-auto-qa-pr` without a PR number.
### 6. Stage the handoff
Follow the shared `om-harness` stage-only contract. Supply the final implementation `--review-packet` and approved `--host-review`
artifacts even in `standard` mode. Stage only allowlisted
intended files, assert refs and reflogs still match the captured start state,
and require a clean staged diff.
Keep the issue claim while the staged branch awaits human review. Render the
report from `references/report-templates.md` and stop.
### 7. Abort or failure cleanup
After capture, follow the `om-harness` stage-only contract to stop active
processes and invoke `end-run` on abort or terminal failure. Preserve the
guard while repairing or retrying the same run; successful staging ends it.
On `--abort`, or on failure after `om-fix` claimed the issue but before a ready
staged handoff exists, release the whole claim — remove the `in-progress` label
through `unlabel-issue`, remove the harness assignee, and post an abort comment
through `comment-issue` — so the issue is not fenced off from later runs. Never
discard staged work unless the user explicitly asks.
## Rules
- Shared communication, marker, label, and safety rules: `references/rules.md`. They always apply.
- For any non-`standard` profile, show the user the probe readiness table (model, binding, role, ✅ ready / ❌ missing-or-failed) before invoking any model.
- After every council round, paste the reviewer-status table and the findings-by-model matrix from the generated `review-summary.md` verbatim into the conversation; the final report repeats the final round's tables.
- Never invoke `om-open-pr`, `om-auto-fix-issue`, `om-auto-create-pr`, or another publication-oriented skill.
- Never commit, push, publish, or execute the tracker `create-pr` operation.
- Keep the issue claim only for a successful staged handoff; release it on abort or failed runs.
- Never count a reviewer from the worker's model family as independent confirmation.
- Never run a required Claude review in the conductor's implementation context; inability to create a fresh Claude context is blocking.
- Never weaken tests or validation to obtain a green result.
- Never modify or clean the user's primary-worktree changes.
© open-mercato · MIT · open-mercato/skills