The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data sourced from remote commit messages and pull request descriptions, which could contain instructions designed to influence the agent.
Ingestion points: The skill retrieves external data using git log, git show, and gh api as seen in SKILL.md (Step 3).
Boundary markers: No specific delimiters or instructions to ignore embedded commands are used when interpolating this external content into the agent's context.
Capability inventory: The skill has access to execute git and gh commands in the user's terminal.
Sanitization: There is no evidence of filtering or sanitization of the fetched commit messages or PR bodies before they are presented to the agent for analysis.
[COMMAND_EXECUTION]: The skill uses shell execution to run git and gh commands. It dynamically constructs command lines using shell variables ($BEFORE, $AFTER) derived from the repository's commit history. This is a standard pattern for git-related tools but constitutes a surface for shell interaction.

pulsum