p-git-commit
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill instructs the agent to bypass standard permission checks by immediately requesting higher system privileges (such as sudo) for Git operations affecting the .git directory, rather than attempting them with current permissions first. Evidence found in SKILL.md: '执行这些命令时应直接申请更高权限，不必先用默认权限试错。'
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from repository diffs and possesses the capability to commit and push changes, which could allow malicious instructions in the code to influence agent behavior.
- Ingestion points: The agent reads repo state via 'git status --short', 'git diff --stat', and 'git diff' as defined in SKILL.md.
- Boundary markers: The skill includes a manual confirmation guardrail, instructing the agent to only commit or push when the user explicitly requests it.
- Capability inventory: The skill executes 'git add', 'git commit', 'git pull', and 'git push' commands.
- Sanitization: No sanitization or filtering of the diff content is mentioned or implemented.
- [COMMAND_EXECUTION]: The skill makes extensive use of Git CLI tools to inspect and modify the local repository.
- Evidence: Shell commands for git status, diff, add, commit, pull, and push are specified in the '工作流' and '提交' sections of SKILL.md.
Audit Metadata