The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources (Instagram comments and direct messages) which can be processed by the agent.
Ingestion points: Untrusted data enters the agent context via scripts/comments.py and scripts/messages.py.
Boundary markers: No specific delimiters or warnings to ignore embedded instructions were found in the scripts or documentation.
Capability inventory: The skill possesses significant capabilities, including extensive network operations (Instagram and Imgur APIs) and local file system access (SQLite database operations).
Sanitization: There is no evidence of content sanitization or filtering for the data fetched from Instagram before it is provided to the agent.
[DATA_EXFILTRATION]: To facilitate content publishing, the skill automatically uploads local media files to Imgur, a well-known third-party hosting service. This is a functional requirement of the Instagram Graph API, which necessitates a public URL for media ingestion. This behavior is documented in the skill's overview and publishing guides.
[METADATA_POISONING]: There is a discrepancy between the author name 'renat' listed in the SKILL.md frontmatter and the author identity 'sickn33' provided in the system context. Additionally, the skill's risk level is self-labeled as 'critical' in the frontmatter, which might be a subjective assessment of the skill's power rather than its security risk.

instagram