tuning-oak-query-indexes
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as
ripgrep(rg) andgrepto scan local files for JCR queries. It also utilizescurlto interact with AEM administrative endpoints (e.g.,/system/console/status-oak-index-defn.json) and perform repository-mutating operations via the Sling POST servlet. - [INDIRECT_PROMPT_INJECTION]: The skill processes queries and index definitions extracted from the target codebase as data for analysis.
- Ingestion points: Source code files analyzed for JCR/QueryBuilder queries (via
references/extracting-jcr-queries.md) and administrative JSON dumps of index definitions. - Boundary markers: The skill instructions provide a methodology to separate static field analysis from dynamic input (Section A vs Section B in
SKILL.md). - Capability inventory: The skill has the capability to read and write to the AEM repository via administrative APIs and execute diagnostic queries through
explaincommands. - Sanitization: The skill emphasizes the use of diagnostic planning commands over direct execution and explicitly mandates user approval before any write operations.
- [COMMAND_EXECUTION]: The skill accesses system-level configuration data, including the full inventory of Oak index definitions via the Felix console, which contains sensitive architectural information about the target application's data model.
Audit Metadata