The Agent Skills Directory

[CREDENTIALS_SAFE]: The protocol contains strict rules against handling sensitive information. It explicitly prohibits touching .env files, wp-config.php, SSH keys, and platform-specific configuration files like Shopify's settings_data.json. It mandates that all credentials remain in the environment and are never included in commands, logs, or output.
[COMMAND_EXECUTION]: The skill performs git operations and executes local scripts via node. Security is maintained by requiring a one-shot, time-limited 'ticket' for sensitive operations and enforcing literal file paths for script execution to prevent injection attacks. The skill intentionally omits Bash and Write tools from its pre-approved list to ensure the user must manually approve every command and file modification.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from plan.json and manifest.json. To mitigate risks, the protocol dictates that it only processes changes explicitly confirmed by the user. It also utilizes the platform's native diffing tools to present every proposed change to the user before it is applied to the filesystem.
[DATA_EXFILTRATION]: There is no evidence of unauthorized data transmission. The protocol forbids the use of general-purpose network tools like curl or wget for remote targets, requiring instead the use of audited local adapter scripts that are subject to internal write guards.

seo-fix-apply