The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes spawnSync to execute local git commands, including git log, git rev-list, and git diff. These commands are used to identify recent commits and summarize file changes since the last saved session to provide the agent with project activity context.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from project files such as CLAUDE.md, package.json, and README.md (ingestion points). Although the skill wraps this data in markdown code blocks and ASCII borders (boundary markers) before presenting it to the agent, it lacks explicit sanitization of the file content. Since the skill grants the agent capabilities to execute local commands and write files (capability inventory), malicious instructions hidden in a project's goal or description sections could theoretically influence agent behavior.

ck