The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The documentation identifies platform-level sensitive headers (X-ZC-PROJECT-SECRET-KEY and X-ZC-Admin-Cred-Token) present in the job runtime context and warns against logging them. It also highlights that the platform's GET APIs return job parameters in plaintext, advising users to avoid storing secrets in those fields to prevent exposure.
[COMMAND_EXECUTION]: The provided Node.js job handler template demonstrates safe handling of input parameters. It uses parseInt() to sanitize values retrieved from the jobRequest before incorporating them into ZCQL (SQL-like) queries, effectively mitigating potential injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill uses the official zcatalyst-sdk-node package and references standard Zoho Catalyst platform components. It does not attempt to download or execute code from untrusted remote sources.
[INDIRECT_PROMPT_INJECTION]: The skill describes an attack surface where untrusted data enters the agent context via job parameters, but it provides guidance for mitigation.
Ingestion points: External data enters the handler through jobRequest.getAllJobParams() (found in SKILL.md and references/job-scheduling-basics.md).
Boundary markers: The documentation explicitly instructs users to "validate/whitelist anything you accept via params" before use.
Capability inventory: The job functions have data access capabilities via app.zcql().executeZCQLQuery, app.datastore(), and the ability to schedule further work via js.cron().createCron().
Sanitization: The provided code templates demonstrate proper sanitization, such as using parseInt() on numerical parameters before use in database queries.
[PRIVILEGE_ESCALATION]: The skill clearly explains the platform's security model, noting that job functions run with admin-level scope for data access because user tokens are not present in background runtimes. This is presented as an architectural requirement rather than an unauthorized escalation of privileges.

catalyst-job-scheduling