The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is instructed to actively scan the repository and environment for sensitive credentials and authentication mechanisms. Specifically, the 'External dependencies' audit searches for credentials needed by SDKs and databases, while the 'Elevation' audit specifically targets keychain and credential-store access.
[PRIVILEGE_ESCALATION]: The skill probes the execution environment for privilege escalation paths. It identifies actions requiring sudo, root rights, or UAC prompts and includes a 'Drill' step designed to execute specific actions to trigger and confirm these elevation prompts.
[COMMAND_EXECUTION]: The skill executes arbitrary shell commands derived from the project's environment and configuration files. During the 'Drill' step, it runs 'validation recipes' and 'smoke e2e' paths which are commands extracted from the repository's CI configuration and task runners.
[DYNAMIC_EXECUTION]: The skill dynamically generates and executes commands at runtime based on its discovery of the project's setup. This includes executing UI automation tools and accessibility drivers to launch and interact with applications.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface as it ingests and acts upon untrusted data from the repository being audited.
Ingestion points: Reads repository files such as CI configs, task runners, agent documentation (CLAUDE.md, AGENTS.md), and project specs.
Boundary markers: None identified; the instructions direct the agent to directly process and execute content derived from these files.
Capability inventory: The skill possesses the ability to execute arbitrary shell commands, dispatch sub-agents, and perform network probing.
Sanitization: There is no evidence of sanitization or validation of the commands parsed from the repository files before they are executed in the 'Drill' phase.

preflight