The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes the !command`` syntax within SKILL.md to execute shell commands and populate the agent's context immediately upon loading.
Evidence: The skill executes git status --short, git diff --cached, git diff, git branch --show-current, and git log -5 --format='%h%n%s%n%b%n---'.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from git diff and git log, which may contain untrusted content from external contributors or malicious code changes.
Ingestion points: Data enters the context via the dynamic execution of git commands in SKILL.md.
Boundary markers: No delimiters or specific instructions are provided to the agent to ignore potential instructions embedded within the source code diffs or previous commit messages.
Capability inventory: The skill empowers the agent to perform git stage and git commit operations based on the analyzed content.
Sanitization: There is no evidence of sanitization or filtering of the ingested git output before it is processed by the agent.
[COMMAND_EXECUTION]: The skill executes local shell commands and a bundled bash script (scripts/loc.sh) to perform its primary functions.
Evidence: SKILL.md instructs the agent to create commits and verify them using git commands, and the scripts/loc.sh utility is provided to calculate lines of code changes for the commit message body.

commit-with-context