The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute shell commands during initialization to install and setup dependencies. Specifically, it executes npm install -g agent-browser and agent-browser install at the start of a session.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs an external dependency (agent-browser) from the NPM registry. This package is used for browser automation and data extraction from arbitrary websites.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes untrusted data from multiple external sources.
Ingestion points: Processes content from Twitter replies, long-form articles, GitHub repository source code, and arbitrary website text (referenced in SKILL.md under Step 2).
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the orchestration flow.
Capability inventory: The skill can execute subprocesses (agent-browser, node), perform network requests to its proprietary API (crab-skill.opsat.io), and write files to the local file system (~/.crab-catch/reports/).
Sanitization: No sanitization or validation logic is defined for the external data before it is passed to the analysis and reporting phases.

Crab Catch