The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process content from external LinkedIn posts and comments. This creates a surface for indirect prompt injection if a post contains instructions designed to manipulate the agent.
Ingestion points: SKILL.md defines steps to "Fetch post body" using lib.ApifyClient.fetch_post(url) and process "untrusted content" from posts, comments, and engager profiles.
Boundary markers: The skill includes a dedicated "Untrusted content" section with explicit instructions for the agent to treat fetched text as data only and never as instructions.
Capability inventory: The skill uses ApifyClient for network fetches and processes text for classification.
Sanitization: Prompt-level instructions are present to prevent the agent from executing commands found within fetched data.
[DYNAMIC_EXECUTION]: The classification logic described in references/classification-rules.md suggests the use of a rule-matching engine to score formulas.
Evidence: The score_formula function in references/classification-rules.md utilizes an eval_feature function to process rules such as anaphora_count >= 3. While these rules are defined within the skill's static configuration files, the pattern involves runtime evaluation of logic strings.

linkedin-hook-extractor