ot-init
Pass
Audited by Gen Agent Trust Hub on Nov 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes two local bash scripts,
on-track-init.shandon-track-index.sh, to manage the repository environment. These scripts perform directory creation and file updates (such as modifyingAGENTS.mdandINDEX.md) within the target project. The execution is limited to local files and does not perform network operations. - [INDIRECT_PROMPT_INJECTION]: The
on-track-index.shscript parses the frontmatter and headings of all markdown files within theon-track/hierarchy to generate a summary index. - Ingestion points: The script reads from markdown files in
on-track/tracks/,on-track/refs/,on-track/research/, and other subdirectories in the target repository. - Boundary markers: While it identifies frontmatter using standard
---delimiters, it does not apply specific sanitization or instructions to ignore embedded commands within the extracted data. - Capability inventory: The skill possesses the capability to read any file in the project's documentation tree and write to the central
INDEX.mdandAGENTS.mdfiles. - Sanitization: Values extracted from file metadata (such as titles, statuses, or slugs) are interpolated directly into a markdown table in
INDEX.mdwithout filtering. An attacker who can commit files to the repository could inject malicious instructions into these fields, which the agent would then ingest when it reads the index to determine its next steps.
Audit Metadata