Skip to content

feat: security review gate for agent-generated PRs #78

Description

@BnJam

Summary

The technocore review pipeline should automatically detect and close PRs containing
malicious code before a human ever sees them. This is particularly important for
agent-generated PRs where the implementing agent could theoretically be prompt-injected
via malicious issue content.

Patterns to detect

  • Credential exfiltration: tokens/keys written to files, env vars sent to external hosts
  • Backdoors: reverse shells, remote eval/exec of downloaded content
  • Supply chain attacks: lockfile tampering, non-canonical registries, typosquatted deps
  • CI/CD tampering: workflow changes that add unexpected secret access or external runners
  • Obfuscated code: heavily encoded strings, steganographic payloads

Proposed behaviour

  1. Reviewer agent runs a security gate before any code review
  2. On detection: close the PR, add security/malicious label, open a security issue
  3. On clean: proceed with standard correctness/maintainability review

Context

This is implemented in the eo-reviewer agent persona in technocore. This issue tracks
the feature so it can be reviewed, refined, and extended as real cases emerge.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions