The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted data from the repository environment, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Reads git commit messages via git log, code changes via git diff, and documentation files including PLAN.md, CHANGELOG.md, and HANDOFF.md.
Boundary markers: The extracted content is interpolated directly into the summary output without delimiters or specific instructions to ignore embedded commands.
Capability inventory: Across all scripts, the skill's capabilities are limited to executing read-only git commands and reading file content; it cannot perform network operations or file-system modifications.
Sanitization: No sanitization or validation is applied to the data retrieved from git or local files.
[COMMAND_EXECUTION]: Executes standard git utilities (git log, git status, git branch, git diff) to gather session context. These commands are executed as local subprocesses and are limited to metadata retrieval within the current working directory.

resume