The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from a user-defined scope, such as branch diffs and untracked files, to perform its adversarial review.
Ingestion points: Branch diffs, working tree content, and untracked files identified via git ls-files.
Boundary markers: The skill uses a brief for subagents but does not specify delimiters or warnings to ignore embedded instructions in the reviewed code.
Capability inventory: Spawns subagents (Rex workers), executes repro scripts, and runs shell commands (git).
Sanitization: No explicit sanitization or escaping of ingested content is mentioned before it is processed or executed by subagents.
[COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary commands to verify bugs.
Evidence: Subagents are instructed to "Prove each finding with a run against the real code, such as a failing test, a repro script, or command output".
Evidence: The skill uses shell commands like git diff HEAD --binary > /tmp/interrogate-<slug>.patch and git ls-files --others --exclude-standard to gather data for analysis.

interrogate