You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Adds the omv CLI so users can install skills via npx oh-my-vul setup.
Updates README, CLAUDE.md, CI workflow, and package metadata.
Constraint: zero runtime dependencies — only @types/node + typescript as devDeps
Confidence: high
Scope-risk: narrow
This repository contains a small multi-skill project for passive vulnerability research and VulDB/CVE reporting.
3
+
oh-my-vul is a modular, community-oriented collection of LLM-friendly vulnerability research skills for Claude Code.
4
4
5
-
The project is mostly Markdown instructions plus deterministic helper scripts. There is no compiled application code.
5
+
The project ships a TypeScript CLI (`omv`) for installing skills, plus Markdown skill definitions and Python dev tools. Users install via `npx oh-my-vul setup`.
6
6
7
7
## Structure
8
8
9
-
-`vuln-finder/SKILL.md` - skill entry point for finding and ranking open-source packages worth auditing.
Packages should contain root-level `SKILL.md` plus the skill's `references/`, `scripts/`, and `evals/` directories when present. They should not contain nested skill directory entries such as `vuln-finder/` or `vuldb-report/`.
139
+
Packages contain root-level `SKILL.md` plus the skill's `references/`, `scripts/`, and `evals/` directories when present. They must not contain nested skill directory entries such as `omv-find/` or `omv-report/`.
140
+
141
+
## Contracts
142
+
143
+
Skills reference `contracts/` directly rather than duplicating schema in their `references/`. When adding a new skill that consumes Evidence.v1, reference `../../contracts/evidence.v1.yaml` from the SKILL.md body. The validator handles `../../shared/references/` and `../../contracts/` path prefixes.
90
144
91
145
## Design Notes
92
146
93
-
Keep each `SKILL.md` concise. Put ecosystem-specific guidance, vulnerability matrices, scoring details, examples, and output contracts in `references/` so the agent can load only what the current request needs.
147
+
Keep each `SKILL.md` concise. Put ecosystem-specific guidance, vulnerability matrices, scoring details, examples, and output contracts in `references/` so the agent loads only what the current request needs.
148
+
149
+
When adding new ecosystems or vulnerability classes, update the relevant shared reference file and add at least one eval scenario. Prefer deterministic script checks for repetitive behavior instead of expanding the main prompt.
94
150
95
-
When adding new ecosystems or vulnerability classes, update the relevant reference file and add at least one eval scenario. Prefer deterministic script checks for repetitive behavior instead of expanding the main prompt.
151
+
When adding a new skill: place it in `skills/<name>/` with a `SKILL.md` whose frontmatter `name` matches the directory name, add focused `references/` and `evals/`, then run `python3 scripts/validate_skill.py` and rebuild packages.
0 commit comments