The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script leads.py uses importlib.util to dynamically load and execute a Python module from a relative path (../pipeline/deals.py). This allows the skill to interface with a separate pipeline management component but involves executing code from a path computed at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection because it ingests and processes untrusted data from the open web.
Ingestion points: Raw HTML content is fetched from arbitrary external websites via urllib.request in leads.py.
Boundary markers: None. The skill does not implement delimiters or safety warnings for the agent when it is instructed to "Read the site yourself" in Step 5 of SKILL.md.
Capability inventory: The skill can perform file writes to closer/leads.json, modify the closer/pipeline.json database, and draft or send emails through a Gmail connector.
Sanitization: None. While the script uses regex to identify specific "gaps" (like missing booking links), the agent is encouraged to browse the raw site content, which could contain malicious instructions designed to influence the agent's behavior.
[EXTERNAL_DOWNLOADS]: The audit and rank commands in leads.py perform outbound network requests to fetch content from business websites across the internet. These operations target non-whitelisted domains, although they are central to the skill's stated purpose of auditing leads.
[COMMAND_EXECUTION]: The skill's primary interface involves executing a local Python script (leads.py) via the shell with various flags and arguments, as demonstrated in the usage examples in SKILL.md.

prospector