The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute git commands for repository synchronization and conflict resolution.
[DATA_EXFILTRATION]: The skill executes git push to upload local file contents to a remote repository, which is an intentional transfer of local data.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection when reading and summarizing file contents during git diff or conflict resolution. Evidence: (1) Ingestion: The agent reads file contents in Step 1 and Step 3 of the workflow defined in SKILL.md; (2) Boundaries: No explicit boundary markers or instructions to ignore embedded content are used; (3) Capability inventory: The agent has access to Bash, Read, Write, and Edit tools; (4) Sanitization: No content validation or sanitization is performed on the data read from files before it is processed by the model.

cfg