The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill's setup instructions require users to pass their Caido Personal Access Token (PAT) as a command-line argument: npx tsx caido-client.ts setup <your-pat>. This practice is insecure as it leaves sensitive credentials in the shell's command history and makes them visible in process monitors to other users on the system.\n- [PROMPT_INJECTION]: The skill is highly susceptible to indirect prompt injection. It is designed to retrieve and process raw HTTP history, including requests and responses, from the Caido proxy. This external data is untrusted and could contain malicious instructions designed to manipulate the agent's behavior.\n
Ingestion points: caido-client.ts search, recent, get, and get-response commands that fetch raw HTTP traffic.\n
Boundary markers: The instructions do not specify any delimiters or safety warnings to ensure the agent ignores embedded instructions within the proxy data.\n
Capability inventory: The skill can execute shell commands (npx tsx), perform network operations via the Caido SDK, and manage sensitive secrets.\n
Sanitization: There is no evidence of sanitization or filtering of the retrieved HTTP content before it is processed by the agent.\n- [DATA_EXFILTRATION]: The skill accesses and manages sensitive data, including authentication tokens and environment variables (victim IDs, tokens, etc.), which it persists to the local file ~/.claude/config/secrets.json. This represents an exposure risk for sensitive testing metadata and credentials.\n- [COMMAND_EXECUTION]: All functionality in the skill relies on executing shell commands through npx tsx to interact with a local TypeScript client. While standard for this architecture, it creates a broad surface for command-based interaction with the system.\n- [EXTERNAL_DOWNLOADS]: The skill documentation identifies a dependency on the @caido/sdk-client package. While this appears to be a legitimate library for the Caido platform, it is an external dependency that the skill interacts with.\n- [NO_CODE]: The core logic of the skill resides in a file named caido-client.ts and a lib/ directory which are not included in the provided skill bundle. The analysis is based on the instructions and command descriptions provided in the SKILL.md file.

caido-mode