The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text data from external sources (X/Twitter, Web Archive, and Common Crawl) to perform its primary function of post recovery.
Ingestion points: External data enters the agent context via scripts/search_twitter.py, which fetches content from fxtwitter.com, vxtwitter.com, web.archive.org, and commoncrawl.org.
Boundary markers: SKILL.md (Step 6) explicitly requires the agent to deliver results with clearly labeled sections for verbatim text, OCR transcriptions, and editorial notes to prevent source confusion.
Capability inventory: The skill uses urllib.request for network operations and file writing via scripts/profile_store.py for session-persistent settings.
Sanitization: The extraction script uses HTML unescaping and regular expression filtering to clean retrieved data, while the system instructions enforce strict verbatim reporting.
[DATA_EXPOSURE_&_EXFILTRATION]: The skill manages user preferences through scripts/profile_store.py, which includes a security check to block the storage of sensitive keys such as tokens, secrets, or passwords. It exclusively reads/writes to defined configuration directories for state management.
[COMMAND_EXECUTION]: All primary operations are performed by Python scripts using the standard library. The scripts do not use dangerous functions like os.system or subprocess.run to execute user-provided strings.

lov-search-twitter