codify
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted text from external sources that can be controlled by third parties.
- Ingestion points: Section 3 of
SKILL.mdspecifies that the agent gathers data from human comments (conversation, review, inline), review findings in PR bodies, and commit messages. - Boundary markers: The skill does not define delimiters or specific instructions to ignore malicious directives embedded within the review comments it processes.
- Capability inventory: The skill has the ability to write to the filesystem (modifying
.agents/coding-standards.md) and execute shell commands (running the project's check command and git operations). - Sanitization: There is no evidence of sanitization, escaping, or validation of the text gathered from PRs before it is used to generate new coding rules or influence the 'check command' execution.
- [COMMAND_EXECUTION]: The skill performs shell-level operations that depend on the project environment.
- Evidence: Section 7 of
SKILL.mdinstructs the agent to 'run the project's check command once' after a configuration change. Since this command is not statically defined in the skill, its behavior depends on local project configurations which could be tampered with. - Evidence: Section 8 of
SKILL.mdusesgit status --shortandgit check-ignoreto report the state of the repository.
Audit Metadata