The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): The skill requires the input of Hive private keys (posting and active) via the command line and stores them in ~/.hive-cli/config.json. This exposes sensitive credentials to shell history and system logs.- [PROMPT_INJECTION] (HIGH): High risk of indirect prompt injection. Malicious content from the Hive blockchain, ingested via hive content or hive call, could contain instructions that trick the agent into using its hive transfer or hive broadcast capabilities to steal funds or post malicious content.
Ingestion points: hive content <author> <permlink>, hive call database_api in SKILL.md.
Boundary markers: Absent. No delimiters or safety instructions provided for external data.
Capability inventory: hive transfer (financial), hive vote, hive comment, hive broadcast in SKILL.md.
Sanitization: None. External data is not validated or sanitized before being handled by the agent.- [EXTERNAL_DOWNLOADS] (MEDIUM): The skill installs the @peakd/hive-tx-cli package from an untrusted source (GitHub user asgarth), which may contain unverified code.- [COMMAND_EXECUTION] (HIGH): The skill enables execution of arbitrary blockchain transactions (transfers, custom JSON) which can have significant real-world financial consequences if the agent is compromised via injection.

hive