The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill facilitates the retrieval of sensitive information using the hodman env get and hodman git secret commands with the --unsafe flag. This allows access to environment variables (e.g., API keys) and private keys in the project environment.\n
Evidence: The 'Secret reveal' section in SKILL.md explicitly details the use of the --unsafe flag for these operations.\n
Mitigations: The skill instructs the agent to never print, persist, or summarize these secrets in tasks, threads, or reports.\n- [COMMAND_EXECUTION]: The agent is granted the ability to execute arbitrary shell commands and SQL queries within the project's environment, which is a powerful capability.\n
Evidence: Commands such as hodman shell exec and hodman sql query are documented for use in diagnostic and operational tasks.\n- [DATA_EXFILTRATION]: The skill provides tools for reading project files, querying databases, and accessing environment variables, establishing a capability for data extraction.\n
Evidence: hodman file get, hodman sql query, and hodman env get are core features of the toolset.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes data from untrusted external sources, which could contain malicious instructions designed to influence the agent's behavior.\n
Ingestion points: Project logs (hodman logs), source code via file cat, and incoming messages from external channels like Telegram via thread send (reading the thread history).\n
Boundary markers: The skill does not define specific delimiters or "ignore" instructions for processing these data streams.\n
Capability inventory: The agent has high-privilege capabilities including shell exec, file put, sql query, and env get across the project environment.\n
Sanitization: While safety warnings are provided regarding credential handling, there is no technical sanitization specified for content ingested from external sources before it is processed by the agent.

hodman