The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the Farcaster network (such as cast text, user bios, and search results) via the Neynar API. This data is ingested into the agent's context and could theoretically contain instructions designed to influence the agent's behavior.
Ingestion points: scripts/neynar.sh fetches external content using the /feed, /search, and /cast endpoints.
Boundary markers: The script returns data as JSON objects filtered by jq, but does not wrap the natural language fields in explicit boundary markers or safety instructions.
Capability inventory: The skill utilizes curl for network requests to api.neynar.com and reads its own configuration from ~/.clawdbot/skills/neynar/config.json.
Sanitization: The script uses jq -Rr @uri to encode search queries and URL identifiers. However, other identifiers like usernames and FIDs are passed directly, which could allow for minor HTTP parameter injection if not validated by the agent.
[DATA_EXFILTRATION]: The skill transmits data to and from api.neynar.com as part of its core functionality.
Evidence: scripts/neynar.sh uses curl to send API keys and user-provided cast content to the Neynar API base URL.
[COMMAND_EXECUTION]: The skill executes standard system utilities to perform its tasks.
Evidence: SKILL.md and scripts/neynar.sh rely on curl for network operations and jq for robust JSON processing.
[SAFE]: The skill implements secret management best practices by instructing users to place sensitive API keys and signer UUIDs in a dedicated config.json file rather than hardcoding them in scripts.
[SAFE]: For write operations like posting or reacting, the script uses jq --arg to assemble JSON payloads, which prevents JSON injection attacks by ensuring all user input is properly escaped as string literals.

neynar