Skip to content
Back to skills

Vcsdd Feedback Routing

ASecurity

Use this skill when routing adversary findings in Phase 4. Provides routing logic, iteration limit enforcement, and escalation protocols.

  • 27 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added June 1, 2026
code-qualitygosecurity

Security analysis

A100/100

Scanned June 1, 2026

npx -y skills add sc30gsw/vcsdd-claude-code --skill vcsdd-feedback-routing --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Vcsdd Feedback Routing?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Vcsdd Feedback Routing
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/sc30gsw-vcsdd-feedback-routing/badge)](https://www.skillsdirectory.com/skills/sc30gsw-vcsdd-feedback-routing)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: vcsdd-feedback-routing
description: Use this skill when routing adversary findings in Phase 4. Provides routing logic, iteration limit enforcement, and escalation protocols.
origin: VCSDD
---

# VCSDD Feedback Routing

## When to Activate
- Phase 4 (feedback integration loop)
- Parsing adversary verdict.json
- Deciding which phase to route findings to

## Routing Table

| Finding Category | Severity | Route To |
|------------------|----------|----------|
| spec_ambiguity | any | Phase 1a |
| spec_gap | any | Phase 1a |
| requirement_mismatch | any | Phase 2b |
| missing_edge_case | critical | Phase 1a |
| missing_edge_case | high/medium/low | Phase 2a |
| test_coverage | any | Phase 2a |
| test_quality | any | Phase 2a |
| implementation_bug | any | Phase 2b |
| error_handling | any | Phase 2b |
| security_surface | any | Phase 2b |
| code_structure | any | Phase 2c |
| naming | any | Phase 2c |
| duplication | any | Phase 2c |
| proof_gap | any | Phase 5 |
| invariant_violation | any | Phase 5 |
| purity_boundary | any | Phase 1b by default; Phase 2c or 5 only when the architecture still stands |

**Always route to the EARLIEST affected phase.**

## Iteration Limit Enforcement

| Phase | Max Iterations | On Exceed |
|-------|---------------|-----------|
| 1c | 3 | Write escalation, pause for human |
| 3 (strict) | 5 | Write escalation, pause for human |
| 3 (lean) | 3 | Write escalation, pause for human |
| 6 | 2 | Write escalation, pause for human |

## Escalation Protocol

When iteration limit exceeded:
1. Write to `.vcsdd/features/<name>/escalations/escalation-<timestamp>.md`
2. Include: phase, iteration count, limit, all open findings
3. Surface to human: "Human review required. N findings remain after M iterations."
4. Pause pipeline - do not auto-advance

## Feedback Integration Steps

1. Read `reviews/sprint-N/output/verdict.json`
2. Read all `reviews/sprint-N/output/findings/FIND-NNN.json`
3. Group findings by `routeToPhase`
4. Route to earliest affected phase (e.g., if 1a and 2b both affected, route to 1a first)
5. Update finding beads to status "open" with resolution.status "open"
6. Transition pipeline via explicit Phase 4 routing:
   - if current phase is `3`, enter `4` first
   - from `4`, move to the target phase
   - prefer `routeFeedback(featureName, targetPhase, reason)` over hand-written `transitionPhase()` chains

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.