The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documentation instructs the agent to clone an external repository (github.com/solanatracker/raptor-binary) and execute a binary (./raptor). This pattern of downloading and running arbitrary executable code from an unverified source represents a significant security risk.\n- [EXTERNAL_DOWNLOADS]: Reference files (references/raptor_setup.md) encourage the retrieval of external assets, including a binary file and an authentication 'signature' file, from a repository not belonging to a trusted organization.\n- [COMMAND_EXECUTION]: The setup instructions include shell commands to modify file permissions (chmod +x raptor) and execute the binary with sensitive environment variables such as RPC_URL and PRIVATE_KEY passed in the command line environment.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the SolanaTracker API and interpolates it into analysis output.\n
Ingestion points: scripts/token_analysis.py and scripts/wallet_pnl.py (via httpx calls to data.solanatracker.io).\n
Boundary markers: Absent. The scripts output data directly to the agent context without delimiters or warnings to ignore embedded instructions.\n
Capability inventory: Binary execution (./raptor in references/raptor_setup.md) and network operations (httpx in scripts/token_analysis.py, scripts/wallet_pnl.py).\n
Sanitization: Absent. No validation or filtering is performed on the API response data.

solanatracker-api