The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data by processing CLI outputs, including deployment snapshots, billing ledgers, and runtime logs. This creates a potential attack surface where untrusted data could influence agent actions.
Ingestion points: Processes output from xapi workers pull, xapi workers logs, and generated ledger.json files (SKILL.md, references/billing.md).
Capability inventory: The agent has capabilities to deploy code (workers push), delete resources (workers delete), and manage credentials (secrets set) through the xapi CLI.
Sanitization: The instructions include specific mandates to sanitize evidence by excluding tokens, cookies, and passwords before reporting results (SKILL.md).
[EXTERNAL_DOWNLOADS]: The skill utilizes standard development tools and package managers for building worker applications.
Evidence: Mentions the use of npx wrangler, npm install, and various framework templates like Vite and Next.js, which are standard components of the Cloudflare development ecosystem (references/deployment.md).
[DATA_EXFILTRATION]: The skill identifies a sensitive configuration path for authentication.
Evidence: Mentions ~/.xapi/config.json as a supported source for authentication credentials. This is the legitimate, default configuration file for the xAPI CLI tool described in the skill and is required for its primary functionality (SKILL.md).

xapi-workers