The Agent Skills Directory

[COMMAND_EXECUTION]: The skill assembles bash commands using variable interpolation for git queries. * Shell snippets in SKILL.md interpolate user-provided or branch-derived variables like ${SCOPE}, ${ACTION}, and ${BASE_BRANCH} directly into commands such as git log --all --grep="(${SCOPE}". * If the execution environment does not sanitize these variables, an attacker-controlled input could lead to command injection in the host shell.
[INDIRECT_PROMPT_INJECTION]: The skill summarizes external data from git commit logs and diffs which can contain adversarial instructions. * Ingestion points: Data enters the agent's context through the output of git log and git diff commands as specified in SKILL.md. * Boundary markers: Absent; there are no clear delimiters or instructions telling the agent to treat the ingested commit data as potentially untrusted text. * Capability inventory: The agent has the ability to execute local shell commands and read file contents via the git interface. * Sanitization: No filtering or validation is performed on commit messages or diff content before the synthesis step.

recall