fix(tooling): prevent native audit worker recursion

[gfyrag]

Summary

• define audit and challenge providers as inner leaf workers in their existing prompts
• forbid nested ai-audit, ai-audit-challenge, and ai-audit-jira orchestration
• document trusted outer-process ownership and add hostile-manifest prompt regressions

Root cause

The provider prompt marked AGENTS.md and docs/technical/agent-context.md as authoritative without first defining the provider as an already-launched leaf worker. Those repository routing docs tell outer agents to invoke the native audit and challenge launchers, so the provider could follow them recursively from its read-only sandbox.

Bugfix evidence

• BEFORE_FIX=BUG_REPRODUCED: focused prompt-capture tests failed on the exact base because neither worker prompt defined or enforced the leaf role.
• TEST_SENSITIVITY=PASS: temporarily removing the prohibition makes both focused regressions fail.
• The zero-finding fixture runs scripts/ai-audit and then scripts/ai-audit-challenge; its fake provider records no nested-launcher attempt.

Validation

• go test -race ./scripts/aiaudit ./scripts/aiauditchallenge -count=1
• bash scripts/check-repo-invariants
• bash scripts/agent-check
• AI_REVIEW_BASE_SHA=3a8eda3636df48873558140161a852393b4adc6e bash scripts/agent-check-pr
• git diff --check
• exact candidate review: APPROVE, zero findings, residual risk LOW

No audit redesign, PR workflow change, runtime/product change, or Jira publication.

[NumaryBot]

✅ Approve — automated review

The worker prompts clearly establish the non-recursive leaf role before repository guidance is loaded, and the added tests cover both audit and challenge prompts. No actionable defects were found.

No findings.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77.24%. Comparing base (3a8eda3) to head (38e7def).

Additional details and impacted files

@@              Coverage Diff              @@
##           release/v3.0    #1895   +/-   ##
=============================================
  Coverage         77.23%   77.24%           
=============================================
  Files               478      478           
  Lines             51258    51258           
=============================================
+ Hits              39591    39596    +5     
- Misses             8232     8234    +2     
+ Partials           3435     3428    -7     

Flag	Coverage Δ
e2e	77.24% <ø> (+<0.01%)	⬆️
scenario	77.24% <ø> (+<0.01%)	⬆️
unit	77.24% <ø> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.