| name | omv |
| description | oh-my-vul local-first vulnerability research project manager. Shows workspace status, next actions, installed omv-* skills, registry info, and delegates .omv findings lifecycle commands. Use when the user types /omv, asks what to do next, or manages local findings. |
omv
oh-my-vul local-first vulnerability research project manager for Codex and Claude Code.
Process first: For any vulnerability research session (find / audit / report / โwhat nextโ), apply using-omv discipline โ evidence before claims, CLI gates before โconfirmedโ or โready to submitโ, prefer campaign + attack-surface cards before bulk hypotheses. Do not invent parallel workflows outside .omv/ + omv CLI.
Commands
/omv list โ list all installed omv-* skills with one-line descriptions
/omv dashboard โ show workspace, active workflow queue, and recent activity
/omv tui โ open the interactive Ink research workspace in a real terminal
/omv start [flags] โ initialize the private workspace and first campaign
/omv eval โ run deterministic stable skill eval checks
/omv campaign โ list local research campaigns
/omv campaign show <id> โ show Campaign scope, lanes, and next action
/omv campaign surfaces propose <id> โ propose attack-surface cards (ๅผ้ข)
/omv campaign surfaces show <id> โ show proposed/selected surface cards
/omv campaign surfaces select <id> --cards <id,id>
โ select which cards become seed hypotheses
/omv campaign seed <id> โ create candidate Evidence hypotheses (selected cards or lanes)
/omv status โ show local .omv workspace status (delegates to omv CLI)
/omv log โ show local workspace activity log (delegates to omv CLI)
/omv next โ show active findings and recommended next actions
/omv audit <id> โ deep-audit a candidate finding (delegates to omv-audit skill)
/omv repro <id> โ guide local reproduction of a finding (delegates to omv-repro skill)
/omv repro init <id> โ create .omv/repro/<id>/ artifact scaffold
/omv review <id> โ review report readiness and recommend the next step
/omv report artifacts <id> โ check report and reproduction artifacts
/omv report provenance <id> โ hash report inputs into a local provenance manifest
/omv sources init <id> โ capture SourceRef.v1 from known Evidence source facts
/omv sources validate <id> โ check SourceRef.v1 and Evidence hash freshness
/omv verification init <id> โ create .omv/verifications/<id>.yaml adversarial review scaffold
/omv verification show <id> โ show adversarial verification status
/omv verification validate <id>
โ validate Verification.v1 and stale Evidence hash
/omv archive <id> --reason <reason>
โ archive an inactive finding (delegates to omv CLI)
/omv restore <id> โ restore an archived finding (delegates to omv CLI)
/omv findings list โ list .omv/findings evidence files (delegates to omv CLI)
/omv findings show <id> โ show one finding's validation state and next action
/omv findings init <id> โ create a finding template (delegates to omv CLI)
/omv findings validate [id] โ validate one or all findings (delegates to omv CLI)
/omv findings promote <id> โ update finding status (delegates to omv CLI)
/omv help โ show this help text
The first word after /omv is treated as the subcommand. Unknown subcommands show the help text.
Skills in This Collection
| Skill | Invocation | Purpose |
|---|
| omv-find | /omv-find | Find and rank open-source packages worth auditing |
| omv-audit | /omv-audit | Deep-audit a candidate finding, prove or disprove the vulnerability |
| omv-repro | /omv-repro | Guide local reproduction, fill evidence.observed_result |
| omv-report | /omv-report | Generate VulDB/CVE/GHSA/OSV advisory reports |
Registry
Collection metadata lives in references/registry.yaml. Read it to show current version, platform requirements, and per-skill produces/consumes fields.
State Directory
.omv/ at the repository root stores campaigns, findings, source references, report provenance, archive metadata, and the rebuildable local workspace index. It is private local research state and should be gitignored. Campaigns live under .omv/campaigns/; active findings live under .omv/findings/; SourceRef.v1 sidecars live under .omv/sources/; inactive findings live under .omv/archive/findings/.
CLI Delegation
When the user invokes campaign, workspace, lifecycle, repro scaffold, artifact check, archive, or restore commands, run the matching omv CLI command with the available shell tool and display its output. Do not implement the behavior manually (do not mkdir, do not move files, do not write YAML directly).
Use omv help, omv help review, omv help findings, omv help repro, or omv help report as the source of truth for exact CLI signatures. For direct aliases:
/omv dashboard -> omv dashboard/omv tui -> omv tui/omv start ... -> omv start .../omv eval ... -> omv eval .../omv campaign ... -> omv campaign .../omv status -> omv workspace status/omv log -> omv workspace log/omv next -> omv dashboard/omv repro init <id> -> omv repro init <id>/omv review <id> -> omv review <id>/omv report artifacts <id> -> omv report artifacts <id>/omv report provenance <id> -> omv report provenance <id>/omv sources init <id> -> omv sources init <id>/omv sources validate <id> -> omv sources validate <id>/omv verification init <id> -> omv verification init <id>/omv verification show <id> -> omv verification show <id>/omv verification validate <id> -> omv verification validate <id>/omv archive <id> --reason <reason> -> omv findings archive <id> --reason <reason>/omv restore <id> -> omv findings restore <id>/omv findings ... -> omv findings ...
If omv is not found on PATH, output: "omv is not installed. Run npm install --global oh-my-vul, then omv setup --platform codex in Codex or omv setup --platform claude-code in Claude Code."
Subcommand reference
- start / campaign init โ
omv start provides the guided human workflow; omv campaign init creates Campaign.v1 directly for automation.
- campaign list/show โ reads Campaign.v1 files directly and reports generic hypothesis lanes.
- campaign surfaces propose
<id> โ writes .omv/campaigns/<id>.surfaces.yaml with deterministic attack-surface cards from the shared pack catalog, filtered by campaign vulnerability classes. Cards are unproven research topics (ๅผ้ข), not findings.
- campaign surfaces show
<id> โ lists card status (proposed / selected / skipped) and next action.
- campaign surfaces select
<id> --cards <id,id> โ marks chosen cards selected and others skipped. Required before seed when a surfaces file exists.
- campaign seed
<id> โ creates only candidate Evidence.v1 hypotheses. If a surfaces sidecar exists, only selected cards are seeded (finding ids like <campaign>-renderer-pipeline). Otherwise generic vulnerability-class lanes are used. Existing .yaml/.yml findings are never overwritten. Never claim seed audited, reproduced, verified, or proved a vulnerability, and never create ThreatMap, repro, verification, report, or PoC artifacts manually.
- dashboard โ prints workspace status, active workflow queue, and recent activity in one view.
- tui โ opens the read-only Ink workspace when stdin/stdout are TTYs. Bare
omv also opens it in a TTY; omv dashboard, omv --no-tui, JSON, pipes, and CI remain plain.
- eval โ runs checked-in deterministic checker/golden pairs through the unified runner. It does not invoke a model or make network requests;
--json and --junit are available for automation.
- workspace status โ prints workspace path, active/archive counts, status counts, and privacy warnings.
- workspace log โ prints the local activity trail for workspace initialization, finding init, promotion, archive, and restore.
- init
<id> โ creates .omv/findings/<id>.yaml from the Evidence.v1 template; default --status candidate. If file exists, CLI errors โ suggest --force.
- list โ prints ID / STATUS / READY / PACKAGE / VULNERABILITY table for every
.yaml in .omv/findings/.
- review
<id> โ runs the unified pre-report readiness review and returns one verdict: ready, needs-repro, needs-audit, needs-verification, or blocked. Use --strict when adversarial Verification.v1 must pass before reporting.
- show
<id> โ prints one finding's package, vulnerability, validation errors/warnings, missing fields, and next action. Use --archived to inspect archived findings.
- validate
[id|path] โ checks required Evidence.v1 fields; exits non-zero on errors. No arg = validate whole ledger.
- promote
<id|path> --status <s> โ updates the status field and re-validates. Valid statuses: candidate, confirmed, blocked.
- repro init
<id> โ creates .omv/repro/<id>/ with standard reproduction artifact files and records suggested evidence.repro_artifacts.
- report artifacts
<id> โ checks .omv/reports/<id>/ and Evidence.v1 reproduction artifact references before final archive.
- sources init/validate
<id> โ records only source facts already present in Evidence.v1 and checks whether its hash is stale. Never describe SourceRef as proof that a remote source is authoritative.
- report provenance
<id> โ hashes Evidence, non-empty report files, and available SourceRef/ThreatMap/Verification/reproduction dependencies into .omv/reports/<id>/provenance.json.
- verification init
<id> โ creates .omv/verifications/<id>.yaml with the current Evidence.v1 SHA-256 for adversarial verifier review.
- verification show
<id> โ summarizes Verification.v1 decision, disagreements, required changes, and stale-hash state.
- verification validate
<id> โ validates Verification.v1 structure and warns when Evidence.v1 changed after review.
- archive
<id> --reason <reason> โ moves a finding to .omv/archive/findings/ and removes it from active workflow views. For --reason reported, confirmed findings reuse omv report artifacts <id> checks; use --strict to block archive when artifacts are missing or empty.
- archive list โ lists archived findings and archive reasons.
- restore
<id> โ moves an archived finding back to .omv/findings/.
Workflow Overview
omv start โ private workspace + detected target + first campaign
omv campaign init โ target, scope, priorities, generic lanes
omv campaign surfaces propose โ attack-surface cards (pack ร class)
omv campaign surfaces select โ choose which hypotheses to pursue
omv campaign seed โ candidate Evidence.v1 for selected cards (or lanes)
โ
/omv-find โ identifies packages / entry points for those surfaces
writes or updates .omv/findings/<id>.yaml (status: candidate)
โ
/omv-audit โ deep-audits the finding: dataflow trace, guard verification,
PoC description, CVSS scoring, dedup search
updates .omv/findings/<id>.yaml (status: confirmed | blocked | candidate)
โ
/omv-repro โ [optional] guides local reproduction when observed_result is unknown
walks user through execution, records observed_result
updates .omv/findings/<id>.yaml (status: confirmed | blocked)
โ
/omv review โ checks Evidence.v1 plus available ThreatMap.v1 / Verification.v1
returns ready | needs-repro | needs-audit | needs-verification | blocked
โ
/omv-report โ reads confirmed finding, generates VulDB/CVE/GHSA/OSV report
then records/checks local provenance with omv report provenance/artifacts
โ
archive โ omv findings archive <id> --reason reported
Each finding uses one of three Evidence.v1 statuses: candidate, confirmed, or blocked.
Use omv dashboard or /omv next as the canonical active queue view after each stage. When the user asks whether a specific finding can be reported, run omv review <id> first and follow its verdict. When the CLI prints a priority value, follow the highest-priority row first unless the user names a specific finding.
