The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing a local Python engine (scripts/last30days.py) and various sub-scripts. It uses subprocess.run to call external binaries like node (for vendored Twitter search) and yt-dlp (for YouTube data).
[COMMAND_EXECUTION]: Implementation plans in docs/plans/2026-02-03-bird-cli-implementation.md describe a function install_bird() that executes npm install -g @steipete/bird, which attempts to perform a global package installation on the host system.
[CREDENTIALS_UNSAFE]: The skill is designed to read and use local browser cookies (Safari, Chrome, Firefox) to authenticate requests to the X (Twitter) GraphQL API. This behavior, while intended for functionality, involves accessing sensitive user session data.
[EXTERNAL_DOWNLOADS]: The skill connects to multiple external APIs, including api.openai.com, api.x.ai, hn.algolia.com, gamma-api.polymarket.com, api.search.brave.com, api.parallel.ai, and openrouter.ai to transmit search queries and retrieve data.
[PROMPT_INJECTION]: The skill exhibits a significant surface for Indirect Prompt Injection.
Ingestion points: It retrieves and processes untrusted data from Reddit threads, Hacker News comments, X posts, and auto-generated YouTube transcripts (orchestrated in last30days.py and youtube_yt.py).
Boundary markers: Output from research scripts uses Markdown headers (e.g., ### YouTube Videos) to separate sources, but the synthesis instructions in SKILL.md encourage the agent to internalize all content.
Capability inventory: The agent has access to Bash, Write, and WebSearch tools, which could be exploited if malicious instructions are processed from the research results.
Sanitization: No explicit sanitization or filtering of the retrieved content is described before it is passed to the agent for synthesis.

last30days