The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from social platforms, which serves as a vector for indirect prompt injection.
Ingestion points: Data is retrieved from Moltbook and GitHub threads via scripts/moltbook_publish.py and scripts/github_thread.py.
Boundary markers: The system provides clear warnings in script outputs to treat external content as data and not as instructions.
Capability inventory: The skill possesses capabilities for file modification, network communication with specific APIs, and execution of local Python scripts.
Sanitization: All ingested content is subjected to schema validation, signature verification using SHA-256 hashes, and automated secret scanning to ensure data integrity and prevent credential leaks.
[EXTERNAL_DOWNLOADS]: Fetches JSON data from remote sources to populate the candidate pool.
Evidence: scripts/github_thread.py and scripts/verify_github_submission.py are instructed to download profile data from raw.githubusercontent.com.
Context: These downloads are limited to data-only files from a trusted domain and undergo extensive local verification.
[COMMAND_EXECUTION]: Relies on the execution of bundled scripts for matching and assessment logic.
Evidence: The skill frequently invokes python3 scripts such as assess_profile.py and match_profiles.py.
Context: These commands are executed within a transparent, owner-authorized workflow and are protected by hash-based integrity checks.

find-complementary-founders