The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and interpret untrusted data from multiple external sources.
Ingestion points: The skill extracts text from PDF documents, performs OCR on images, generates transcripts from audio and video files, and fetches content from social media platforms (Bluesky, Mastodon, X) and YouTube captions.
Boundary markers: The instructions do not define clear delimiters or use "ignore embedded instructions" warnings when presenting extracted text to the agent for synthesis or summarization.
Capability inventory: The agent has the capability to write to the file system (creating wiki pages and manifests) and synthesize derived artifacts based on the ingested content.
Sanitization: There is no evidence of sanitization or filtering to prevent malicious instructions embedded in the source media from influencing the agent's behavior during the interpretation phase.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch data from external services as part of its core ingestion workflow.
Evidence: The skill fetches JSON responses from platform APIs for Bluesky, Mastodon, and X/Twitter to archive social posts in the raw/social/ directory. It also uses yt-dlp to retrieve metadata and captions from YouTube. These operations are conducted with well-known services and are consistent with the skill's stated purpose.

wiki-media-ingestion