Security audit
cann-review
Security checks for vulnerabilities and agentic risk
Overview
The skill's code and instructions generally match its stated purpose (automatically reviewing GitCode PRs), but there are important inconsistencies and a risky artifact (hard-coded example token in docs/examples) and the package metadata fails to declare the required GitCode API token—you should inspect and confirm before using it with real credentials.
What to check and do before installing or running this skill: 1) Verify the token: Treat the token shown in the examples as suspicious. Search the repository for any other hard-coded secrets and confirm the example token is not valid. If you accidentally used it, rotate/disable the token immediately. 2) Inspect the scripts: The package contains many shell/Python scripts that will run locally and call https://api.gitcode.com. Review gitcode-api.sh and the comment-posting scripts to ensure no unexpected remote endpoints or obfuscated behavior exist. 3) Minimum permissions: When creating a GitCode token for this skill, grant the minimum scope required (the docs suggest 'api' and 'write_repository' for posting comments). Consider a token specifically scoped to only the repos the skill must access. 4) Store credentials securely: Use the config/gitcode.conf file with file permission 600 or set GITCODE_API_TOKEN as an env var in a secure runtime. Avoid placing tokens in global TOOLS.md or sharing them. 5) Test in dry-run / limited mode: Use provided test-api.sh or run scripts with dry-run or on a small test repository before enabling cron automation. 6) Cron and persistence: The skill can be scheduled to run periodically and will post comments automatically. Only enable cron if you intend automatic posting; otherwise run manually. 7) Registry metadata mismatch: Note that the skill registry entry does not list required env vars; treat this as a packaging oversight. Consider asking the maintainer to update the metadata to declare GITCODE_API_TOKEN and any notification env vars. If you are not willing to audit the scripts yourself, run this skill in an isolated/test environment first and avoid using high-privilege tokens. If you find any other hard-coded secrets after inspection, do not use the skill and report it to the publisher or revoke the token(s).
Static analysis
No suspicious patterns detected.