The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses sensitive authentication tokens (GLEAN_INDEXING_TOKEN, GLEAN_CLIENT_TOKEN) from the environment and reads local application logs from /var/log/glean-connector/*.log. While the script attempts to redact token values in the diagnostic manifest, these credentials and potentially sensitive log entries are bundled into a local archive that could be accessed or transmitted.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external Glean API responses and local system logs, creating a surface for indirect prompt injection if the ingested data contains malicious instructions targeting the agent.
Ingestion points: The agent context is populated with content from datasource-config.json, indexing-status.json, search-test.json, and app-logs.txt.
Boundary markers: No explicit delimiter or XML-style tags are used to isolate collected log content or API results from the agent's primary instructions.
Capability inventory: The skill uses Bash for network operations (curl) and file system manipulation, Read for file access, and Grep for filtering.
Sanitization: The skill relies on natural language instructions for the agent to redact sensitive data before final output; the shell script itself performs no automated sanitization of the data retrieved from the APIs or logs beyond redacting the existence of environment variables in a text file.

glean-debug-bundle