The safeguards map / 3.2 Agent harnesses
Sondera Coding Agent Hooks
Drafted with AI
Drafted with AI from the cited sources under the direction of UnGovr staff. UnGovr holds editorial responsibility for this page. How this site is made
Hooks for coding agents such as Claude Code, Cursor and Copilot that check each shell command, file operation and web request against Cedar policies and return allow, deny or escalate.
Where it sits
3.2 Agent harnesses (primary): Receives each shell command, file operation and web request from a coding agent's hooks, evaluates it against Cedar policies before it runs, and returns allow, deny or escalate, with enforcing hooks blocking when the decision service cannot be reached.
A reference monitor for AI coding agents.
captured
returns Allow, Deny, or Escalate. If the harness cannot be reached, enforcement hooks fail closed.
captured
A Cedar forbid here prevents the command from running.
captured
Claude's `Bash` tool, Cursor's shell execution hook, Copilot's `bash` tool, and Gemini's `bash` tool all normalize to the same `ShellCommand` action.
captured
policies intercept every shell command, file operation, and web request
captured
Facts
- Concern
- security
- Names the law
- no
- Subject
- agent
- Used by
- those who run it
- Runs
- at runtime
- Policy comes from
- vendor defaultsowner-entered settings
- Can the agent bypass it
- depends
Handoffs
consumes Policy: Cedar policies
Listed, not drawn: it is not yet generally available.
Every quote is verbatim from a stored capture of the page linked under it. Placement is our reading of the product's own documentation; no publisher has reviewed or endorsed it. Rows CC BY 4.0. Names, marks and logos belong to their owners; logos identify each product and are not part of the dataset.