The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to dynamically discover and execute command-line tools within the local environment.
The agent is instructed to "Discover local CLIs and their current help rather than assuming a universal launch or login command."
Explicit commands are specified for usage, such as cap guide --json and ffprobe for media inspection.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external repository data to guide its actions, creating a vulnerability surface where malicious instructions in that data could influence the agent.
Ingestion points: The agent is directed to read various repository files including agent guides, feature maps, recipes, and test harnesses (SKILL.md).
Boundary markers: The instructions do not define specific delimiters or guardrails to prevent the agent from obeying instructions embedded within these external documents.
Capability inventory: The skill utilizes shell command execution (via Cap, ffprobe, and Playwright), file system access for managing recording artifacts, and UI interaction capabilities.
Sanitization: The skill encourages manual inspection of the final recording for secrets, but it does not specify sanitization or validation of the input data processed during the recording setup.

record-verification