The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses local sensitive files to manage user credentials.\n
Evidence: The script scripts/symbiosis-swap.py reads the user's API key from ~/.bankr/config.json. This key is used to authenticate requests to the Bankr Submit API to execute transactions.\n- [EXTERNAL_DOWNLOADS]: The skill interacts with third-party service endpoints to fulfill swap requests.\n
Evidence: Requests are made to https://api-v2.symbiosis.finance/crosschain/v1/swap to retrieve quotes and transaction payloads. This is a well-known service but remains an external dependency.\n- [COMMAND_EXECUTION]: The skill relies on the execution of Python scripts to bridge wallet functionality with the Symbiosis protocol.\n
Evidence: The agent is instructed to run scripts/symbiosis-quote.py and scripts/symbiosis-swap.py locally. These scripts use standard libraries to perform network operations and submit blockchain transactions.\n- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to processing untrusted data.\n
Ingestion points: User-supplied token addresses, amounts, and chain IDs derived from prompts and instructions in SKILL.md.\n
Boundary markers: No specific delimiters or warnings are used to sanitize or isolate user-provided inputs.\n
Capability inventory: The skill utilizes bankr_submit in scripts/symbiosis-swap.py to execute transactions on behalf of the user.\n
Sanitization: No explicit input validation or escaping is observed in the provided scripts before interpolation into API payloads.

symbiosis