The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs the agent to dynamically generate and execute custom "helper" scripts (e.g., in Python or Node.js) to manage API requests and credential handling.
Evidence: SKILL.md and references/helper-contract.md outline a requirement for the agent to write a helper script into the project if a suitable one does not already exist.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a significant attack surface by ingesting and processing untrusted data from external sources while maintaining shell and file-writing capabilities.
Ingestion points: SKILL.md lists untrusted inputs including API responses, error bodies, environment files, logs, and general repository files.
Boundary markers: No specific delimiters (like triple backticks or custom markers) are explicitly required for wrapping untrusted input during processing.
Capability inventory: The agent has capabilities for shell command execution and file system writes (writing helper scripts and temporary output files).
Sanitization: The skill provides a robust multi-step procedure to check output text for credentials in various formats (raw, URL-encoded, and Base64) to prevent accidental leaks.
[COMMAND_EXECUTION]: The skill's core functionality relies on executing shell commands to interact with external services and check the local environment.
Evidence: Procedures in SKILL.md involve running shell-based project helpers and using commands like rg or grep to verify the presence of configuration files.
[DATA_EXFILTRATION]: The skill is explicitly designed to handle and transmit sensitive credentials (API tokens, passwords) to external services.
Evidence: The procedures describe reading credentials from environment files (.env) and passing them to external API endpoints through localized processes.

secret-hygiene