The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, untrusted sources including pull request bodies, commit messages, issues, and review comments. This content could potentially contain adversarial instructions aimed at manipulating the agent's logic or bypassing audit rules.
Ingestion points: PR body, commit messages, linked issues, current checks, reviews, and comments.
Boundary markers: None specified in the instructions to separate untrusted PR data from system instructions.
Capability inventory: The agent can build and execute binaries, run shell commands for testing, push code to PR heads, and perform mutations on App Store Connect resources.
Sanitization: No explicit sanitization or filtering of PR content is described before processing.
[DYNAMIC_EXECUTION]: The skill instructions include building binaries from the pull request source code and executing them to verify behavior. This involves the dynamic compilation and execution of code that has not yet been merged or fully vetted.
[COMMAND_EXECUTION]: The skill uses various shell-level operations to perform its tasks, including creating dedicated worktrees, running build tools, executing test suites, and performing CLI invocations against built binaries.

audit-asc-pr