The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from git diffs, merge requests (MR), and pull requests (PR), creating a vulnerability surface where malicious instructions hidden in code comments or commit messages could influence the agent's behavior.
Ingestion points: External data is ingested via GitLab/GitHub URLs or raw diff text as defined in reference.md.
Boundary markers: Absent. The instructions do not define clear delimiters or warnings to ignore instructions embedded within the processed content.
Capability inventory: The skill can execute bash commands (git, glab, gh) and has file-writing capabilities to the local filesystem (Step 5 in SKILL.md).
Sanitization: Absent. There is no explicit instruction to sanitize or escape the content of the diffs before analysis.
[COMMAND_EXECUTION]: The skill uses local CLI tools including git, glab, and gh to retrieve code changes. While these are used for the skill's primary purpose, they involve executing shell commands with arguments derived from potentially untrusted URLs.
[EXTERNAL_DOWNLOADS]: The skill references an external repository FuDesign2008/open-skills for dependency installation. It instructs the agent to prompt the user to run an npx command to install missing skills if a prerequisite check fails.

test-guide-from-code