The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and synthesize large amounts of untrusted data from various external platforms.
Ingestion points: Untrusted data enters the agent context via web_search (Brave), reddit_search, bird search (X/Twitter), and web_fetch (full article retrieval).
Boundary markers: The skill's documented workflow and output templates do not define any delimiters or specific instructions to the LLM to ignore or treat with suspicion any instructions found within the fetched external content.
Capability inventory: The skill utilizes network-enabled search tools and local CLI execution (bird), which could be abused if the agent is manipulated by injected instructions.
Sanitization: There is no evidence of content sanitization, filtering, or instruction-stripping before the gathered data is passed to the synthesis step.
[COMMAND_EXECUTION]: The skill workflow involves executing the bird CLI tool with user-supplied topics as arguments (e.g., bird search "[topic]"). This creates a risk where a malicious or cleverly crafted user query could attempt to perform command injection if the underlying agent or CLI tool does not properly sanitize shell metacharacters.
[EXTERNAL_DOWNLOADS]: The skill requires the external installation of the Bird CLI and recommends the reddit-insights MCP server. These are unverified third-party dependencies that the user must install to achieve full functionality.

last30days