The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands (curl, jq, cast, xdg-open) to fetch market data and interact with the user's browser.
Evidence: Found in SKILL.md steps 0, 3, 5, and 7, as well as references/data-providers.md patterns.
Mitigation: Step 3 specifies mandatory input validation using regex (e.g., for Ethereum addresses and decimal amounts) and explicitly rejects any input containing shell metacharacters to prevent command injection.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources including web searches and cryptocurrency pricing APIs.
Ingestion points: SKILL.md (Step 0, Step 5) and references/data-providers.md fetch data from DexScreener, DefiLlama, and general web results.
Boundary markers: Explicit instructions require the agent to label web-discovered data as untrusted, provide risk warnings, and obtain user confirmation via AskUserQuestion before proceeding.
Capability inventory: The skill has the ability to execute network requests, parse JSON, and open URLs in the local environment.
Sanitization: The skill implements strict regex-based validation for all user-supplied or externally-fetched variables before they are used in commands or link generation.

swap-planner