The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to fetch the @tokenchit/cli package from the npm registry, which is a third-party dependency from an unverified author.
[REMOTE_CODE_EXECUTION]: The package downloaded from the npm registry is executed directly on the host machine using npx to perform initialization and synchronization tasks.
[COMMAND_EXECUTION]: The skill invokes shell commands such as npx init and npx sync to interact with the local environment and generate the output files.
[DATA_EXFILTRATION]: The skill accesses local AI agent session logs and metric files (e.g., ~/.claude/metrics/costs.jsonl). While intended for generating usage statistics, executing unverified code with access to these sensitive activity logs presents a risk of data exposure. The tool also includes an optional publish command to send data to a public leaderboard.
[INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for local logs generated by agent sessions. If these logs contain malicious content or instructions from previous tasks, they could influence the tool's behavior. \n
Ingestion points: AI agent session logs and local metric files (SKILL.md). \n
Boundary markers: None specified to delimit or protect the log parsing process. \n
Capability inventory: Shell execution via npx and filesystem write access for SVG and configuration files (SKILL.md). \n
Sanitization: No explicit sanitization or validation of the log content before processing is documented.
[DYNAMIC_EXECUTION]: The skill relies on the dynamic execution of a Node.js package at runtime using npx. It attempts to mitigate risk by pinning the version to 0.10.0 and suggesting integrity checks.

token-card