The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions provide commands to programmatically read sensitive API credentials from the local filesystem to authenticate with external services.
Evidence: Use of python3 -c to extract api_key from ~/.gooseworks/credentials.json and export it to the environment.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of a global CLI package from the npm registry as a prerequisite for its functionality.
Evidence: npm install -g @orth/cli listed in the Requirements section of SKILL.md.
[REMOTE_CODE_EXECUTION]: The core functionality of the skill is to download and install new logic and instruction sets from a remote repository, which allows for the introduction of new executable code into the agent's environment.
Evidence: The orth skills add <slug> command allows fetching and deploying capabilities from orthogonal.com.
[COMMAND_EXECUTION]: The skill instructions utilize shell commands to manage environment variables, interact with the local filesystem, and inspect downloaded content.
Evidence: Execution of cat ~/.openclaw/skills/<slug>/SKILL.md and use of npx goose-skills in the metadata for installation.
[PROMPT_INJECTION]: The skill reads and processes the content of externally sourced instruction files, creating an entry point for indirect prompt injection if the downloaded skill definitions contain malicious directives.
Evidence: Ingestion points: cat ~/.openclaw/skills/<slug>/SKILL.md. Capability inventory: orth skills add and shell command execution. Boundary markers: Absent. Sanitization: Absent.

find-skill