The Agent Skills Directory

[METADATA_POISONING]: The YAML frontmatter in SKILL.md identifies the author as 'longbridge', which contradicts the verified author context 'nickbuddy2046'. This discrepancy is misleading regarding the skill's origin and vendor identity.
[DATA_EXPOSURE]: The skill accesses sensitive user information, specifically private watchlist data and portfolio positions, through the Longbridge Watchlist API and Longbridge Positions API mentioned in references/catalyst-radar.md. While used for financial analysis, this represents the exposure of private financial records.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external content from stock news, regulatory filings, and web searches, creating a surface for indirect prompt injection.
Ingestion points: longbridge news, longbridge filing, longbridge analyst-estimates, and various market data sources retrieved via WebSearch (e.g., TAM figures, analyst reports).
Boundary markers: The instructions do not implement explicit delimiters or instruction-ignore warnings when interpolating these external data sources into the agent context.
Capability inventory: The skill utilizes a large number of longbridge CLI subcommands to fetch and manipulate data, which could be influenced by malicious instructions embedded in financial reports.
Sanitization: No sanitization or filtering logic is described for the content retrieved from external news or filing sources.
[COMMAND_EXECUTION]: The skill relies extensively on an external CLI dependency (longbridge-terminal) to execute market data commands such as screener, rank, anomaly, constituent, quote, and kline. The reliance on external, non-whitelisted tooling for primary functionality introduces a dependency risk.

stock-intel