RepoJournal
Agent Security Agent Security
16 wires and counting

$ follow Agent Security

Keep up with Agent Security in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.

or

fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.

Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.

$ status

wire 2026-10-15
stories 23

© 2026 RepoJournal Home Showcase How it works Privacy

$ the-wire · showcase

Guardrails adds CEL rules per tool, agent-scan stops missing Copilot CLI projects

By RepoJournal · Filed · About Agent Security · Composed from the cited sources · methodology

Two operator-visible fixes landed: NVIDIA-NeMo/Guardrails gained a deterministic CEL rail that reads tool-call arguments directly, and snyk/agent-scan now finds MCP servers and skills in Copilot CLI projects it previously skipped.

feat: add per-tool CEL expression rail NVIDIA-NeMo/Guardrails

by tanushriya910

NVIDIA-NeMo/Guardrails shipped a `cel` rail that blocks a tool call or tool result when any Common Expression Language expression evaluates to true. Because an expression reads the structured fields of a call rather than matching text, one rule can combine arguments, as in `has(args.cwd) && args.cwd.startsWith("/etc")`, and it calls no model.

fix: ADS-1387 discover Copilot CLI session projects snyk/agent-scan

by SH4DY

snyk/agent-scan read project roots only from `permissions-config.json`, but Copilot CLI also records projects in `~/.copilot/session-state/<id>/workspace.yaml`, so those projects' MCP servers and skills were never scanned. The tool now also collects roots from each session's `cwd`/`git_root` and the trusted-folders list, verified on Linux, macOS and Windows.

fix: require success:true payload for the hooksConfigured test event snyk/agent-scan

by lbeurerkellner

Agent Monitor answers HTTP 200 with an empty body when a push key is rejected, which meant agent-scan's install-time hooksConfigured test event passed with an invalid key. The test now counts as successful only when the hook script's stdout is a JSON object with `"success": true`; a non-zero exit still fails, and other callers of `_invoke_hook_script` keep the old exit-code behavior behind an o...

docs(i18n): match the OWASP badge to the mapping in every README microsoft/agent-governance-toolkit

by imran-siddique

Four translated READMEs still carried the `10/10 Covered` OWASP badge and linked to a mapping that says otherwise, after the English badge was corrected to `7 Full, 3 Partial`. This pass brings the translations in line, so readers of the ja and other localized docs stop getting the wrong compliance picture.

Quick answers

What shipped in Agent Security on October 15, 2026?
Two operator-visible fixes landed: NVIDIA-NeMo/Guardrails gained a deterministic CEL rail that reads tool-call arguments directly, and snyk/agent-scan now finds MCP servers and skills in Copilot CLI projects it previously skipped. In total, 4 commits and 19 pull requests landed.
Who contributed to Agent Security on October 15, 2026?
5 developers shipped this update, including imran-siddique, dependabot, tanushriya910, SH4DY, and lbeurerkellner.
What were the notable Agent Security updates?
feat: add per-tool CEL expression rail, fix: ADS-1387 discover Copilot CLI session projects, and fix: require success:true payload for the hooksConfigured test event.