update-project-rules
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes several shell commands including
git submodule,git status,git diff, andgit committo manage and report on the project's repository state. - [REMOTE_CODE_EXECUTION]: The skill performs a
git submodule update --remoteto fetch code from the vendor's repository and then executes a shell script (.claude/ai-dev-kit/rules/build-agents-md.sh) from the updated submodule. This results in the execution of code retrieved from a remote source. - [INDIRECT_PROMPT_INJECTION]: The skill reads and processes external data, specifically checking
AGENTS.mdfor a marker and displayinggit diffoutput to the user. - Ingestion points: Data is ingested from the file system (
AGENTS.md) and from the output ofgitcommands. - Boundary markers: No explicit delimiters are used when reading file contents or command outputs.
- Capability inventory: The skill has the capability to execute shell commands and shell scripts.
- Sanitization: The skill does not perform sanitization on the file content or git output before processing or presenting it to the user.
Audit Metadata