The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches and installs the mcptools package from an untrusted third-party repository (github.com/f/mcptools).
[EXTERNAL_DOWNLOADS]: Downloads the axiom-mcp tool from Axiom's official GitHub repository.
[CREDENTIALS_UNSAFE]: The skill instructs the agent to search for sensitive AXIOM credentials in local files such as .env and .config, which could expose secrets to the agent's context.
[PROMPT_INJECTION]: Potential for indirect prompt injection from untrusted Axiom log data. Mandatory Evidence Chain: (1) Ingestion points: log entries retrieved via queryApl. (2) Boundary markers: no delimiters or instructions to ignore embedded commands are present. (3) Capability inventory: access to Bash and Write tools provides a significant attack surface. (4) Sanitization: no evidence of filtering or validation of log content.
[COMMAND_EXECUTION]: Executes shell commands via Bash for dependency installation and interacting with external binaries in the ~/go/bin directory.

axiom-audit