The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a significant amount of untrusted data from external sources like news articles, social media, and forums (e.g., Glassdoor, Zhihu) to evaluate management performance and sentiment.
Ingestion points: External data enters the agent context through the WebSearch tool and sub-agents (Task tool) across multiple steps of the research process.
Boundary markers: The instructions do not define explicit boundary markers or delimiters to separate external content from the agent's core instruction set.
Capability inventory: The skill possesses capabilities that could be targeted by an injection attack, including executing shell commands via Bash, writing research reports to the local file system (Write), and spawning additional autonomous agents (Task).
Sanitization: No explicit sanitization or filtering logic is specified for the data retrieved from external web searches.
[COMMAND_EXECUTION]: The instructions require the agent to execute local shell commands to establish context and verify data.
Evidence: The skill explicitly instructs the agent to run the date command for baseline temporal context and to execute a repository-local tool: python3 tools/financial_rigor.py ....
[EXTERNAL_DOWNLOADS]: The skill relies on retrieving data from remote web services to perform its core function.
Evidence: Extensive use of WebSearch to confirm key figures, collect public statements, and gather feedback from employees and customers across multiple investigative phases.

management-deep-dive