The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill downloads and executes the Homebrew installation script using a shell command substitution pattern.
Evidence: The script in SKILL.md executes /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)" during the prerequisites check.
[COMMAND_EXECUTION]: The skill instructions grant the agent autonomous control to execute a wide range of shell commands for system configuration, repository management, and package installation without user confirmation.
Evidence: Includes instructions to run brew install, docker compose, gh repo clone, and various wendy CLI commands.
[INDIRECT_PROMPT_INJECTION]: The skill implements an autonomous loop that ingests data from external sources to diagnose bugs and apply code fixes, which are then committed and merged.
Ingestion points: Processes output from make test-swift (test suite results), docker compose logs (service logs), and the repository source code itself.
Boundary markers: Absent; there are no specific delimiters or instructions to ignore instructions that might be embedded in logs or ingested files.
Capability inventory: The agent possesses high-privilege capabilities including git push, gh pr create, and gh pr merge.
Sanitization: Absent; no filtering, validation, or escaping of the ingested data is defined before it is processed by the agent.
[CREDENTIALS_UNSAFE]: The skill instructions involve reading local temporary files that contain sensitive authentication artifacts to verify session status.
Evidence: The skill uses cat /tmp/wendy-auth-out.txt to monitor for strings such as "Received enrollment token" and "Authentication successful".

wendy-cloud-iterate