The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill provides commands that output sensitive information.
Evidence: echo $INSTANTLY_API_KEY in SKILL.md prints the raw API key to the session logs.
Evidence: env | grep INSTANTLY in SKILL.md exposes environment variables that often contain secrets.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of untrusted error logs.
Ingestion points: Error logs and console messages processed by the agent as described in the diagnostic steps.
Boundary markers: None provided to isolate data from instructions.
Capability inventory: Bash(curl:*), Read, and Grep tools are available to the agent for processing found data.
Sanitization: No validation or sanitization is performed on the ingested log data.

instantly-common-errors