The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several shell commands, including find, grep, ls, git, and head, to perform a deep scan of the local project directory. These commands are primarily used to extract technical metadata, README files, documentation, and source code annotations (e.g., TODO/FIXME) to provide context for the product management workflow.
[EXTERNAL_DOWNLOADS]: The skill is designed to automatically fetch and parse content from URLs provided in the user input ($ARGUMENTS). This behavior constitutes a network operation targeting potentially untrusted external domains.
[PROMPT_INJECTION]: The skill exhibits vulnerability factors for indirect prompt injection:
Ingestion points: External data is ingested from arbitrary URLs extracted from user input in the $ARGUMENTS field (SKILL.md, Section 3).
Boundary markers: No specific boundary markers or delimiters are defined in the instructions to isolate untrusted external content from the agent's core logic or subsequent tasks.
Capability inventory: The skill allows the agent to execute shell-based environment scanning tools and write findings to the local filesystem (e.g., docs/pm-context/collect/).
Sanitization: The instructions do not include any steps for sanitizing, validating, or filtering the fetched external content before it is processed or stored.

pm-collect