The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the open web via WebSearch and WebFetch, as well as potentially attacker-controlled content in master_supporting_docs/.
Ingestion points: The skill reads external content during the 'Search for related work' phase using WebSearch, WebFetch, and reading files in the master_supporting_docs/supporting_papers/ directory.
Boundary markers: There are no explicit delimiters or boundary markers mentioned to separate untrusted web content from the agent's instructions during the initial synthesis.
Capability inventory: The skill possesses the ability to write to the file system (Write tool used for saving reports to quality_reports/) and spawn sub-agents (Agent tool for verification).
Sanitization: The skill implements a 'Post-Flight Verification' (CoVe) protocol which requires a separate sub-agent to verify every claim and citation against source materials in a fresh context. This architectural isolation serves as a significant security and accuracy control.
[EXTERNAL_DOWNLOADS]: The skill actively fetches external research data using WebSearch and WebFetch. This behavior is central to the skill's primary purpose of conducting literature reviews and targeting academic or informational sources.
[COMMAND_EXECUTION]: The skill utilizes a set of standard tools (Read, Grep, Glob, Write, WebSearch, WebFetch, Agent, Task) to perform its research functions. The tool usage is consistent with the described functionality of a research assistant.

lit-review