The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to bypass security monitors that scan for secret expansion in command-line arguments. It explicitly directs the agent to read the PAGESPEED_API_KEY directly from the environment within the script to evade the 'permission analyzer' that blocks secret expansions on the Bash line.- [PRIVILEGE_ESCALATION]: The instructions describe how to subvert the platform's 'read-only' mode. It identifies that the security guard reverts writes to code/config paths but 'preserves memory/ and output/', essentially teaching the agent how to achieve persistent file writes despite environmental restrictions.- [PROMPT_INJECTION]: The skill includes instructions that teach the agent how to bypass platform constraints and 'load-bearing' security guards by exploiting specific directory exclusions in the write protection policy.- [INDIRECT_PROMPT_INJECTION]: The skill audits external websites by fetching and parsing HTML, creating a surface for malicious content to influence agent behavior.
Ingestion points: The scripts/seo-audit.mjs script crawls and parses HTML, metadata, and linked resources from external origins provided in the ${var} variable.
Boundary markers: There are no delimited boundaries or instructions provided to the agent to treat the crawled content as untrusted data or to ignore potential instructions embedded in the HTML.
Capability inventory: The skill possesses capabilities for shell execution (Node.js), file writing (via shell redirection to preserved paths), and external communication (via the ./notify tool).
Sanitization: The instructions do not mention any sanitization, escaping, or validation of the external content before it is parsed and reported.

seo-audit