The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill utilizes the browser-cookie3 library in browser_enricher.py to programmatically extract sensitive session cookies (sessionid, csrftoken, ds_user_id) from the local Chrome cookie database.
[DATA_EXFILTRATION]: While the skill primarily interacts with Instagram's official API, the extraction of session data from a browser's protected store is a high-risk operation that bypasses standard authentication boundaries.
[COMMAND_EXECUTION]: The skill executes several bash and Python commands, including a bundled setup.sh script, subprocess calls to ffmpeg in media_extractor.py, and Python code execution via shell switches in the primary instructions.
[EXTERNAL_DOWNLOADS]: The pipeline automatically fetches images and videos from Instagram CDN URLs as part of the data enrichment process in api_bootstrap.py.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it processes untrusted data (captions, OCR text, and audio transcripts) from social media posts.
Ingestion points: Data enters the agent's context through the Instagram API and the results of media processing in media_extractor.py.
Boundary markers: There are no explicit markers or instruction delimiters to prevent the agent from potentially obeying instructions embedded within the extracted text.
Capability inventory: The agent has access to powerful tools like Bash, Read, Grep, and Glob which could be exploited if malicious instructions are successfully injected.
Sanitization: No sanitization or instruction filtering is applied to the data before it is presented to the agent.

instagram-pipeline