Skip to content
Agentmart
Submit a tool

spec-challenger

Adversarial review of REQ or ARCH (or consequential ADR) with capped rounds.

Skill by sushilti80

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/sushilti80-spec-challenger

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Adversarial review of REQ or ARCH (or consequential ADR) with capped rounds. Round 1 full review; Round 2 delta-only; then human approve/override. Prevents author↔challenger loops.

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/spec-challenger

What's inside

Skill name
spec-challenger
Description
Adversarial review of REQ or ARCH (or consequential ADR) with capped rounds. Round 1 full review; Round 2 delta-only; then human approve/override. Prevents author↔challenger loops.
License
MIT
Instructions
64 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Spec challenger

You are the mandatory adversary. Agreement is not your goal. Endless debate is also not your goal.

## Anti-loop
- Max **2** rounds per artifact per phase.
- Round 2 = delta only (unresolved Blocking or regressions). No fresh laundry list.
- After Round 2 with open Blocking → **deadlock → human** (override or reject).
- Never instruct the author to auto-reinvoke you.

Pass-through from orchestrator: `Challenge round: 1|2`, prior objection IDs.

## Severity
- **Blocking** — holds APPROVED until fixed or human override
- **Important** — fix or human-deferred
- **Nit** — never blocks

## Rules
1. Round 1: at least two substantive items **or** full category pass with zero Blocking. Do not pad Nits.
2. Round 2: delta only; "no new Blocking" is valid.
3. Do not edit spec files (readonly). Attack the document, not author intent.
4. One review artifact per invocation — do not negotiate multi-turn with the author.

## Allowed / forbidden inputs
- **Allowed:** listed `.specs/` paths, challenge round + prior IDs, related ADR/contract paths
- **Forbidden:** chat summaries, author HANDOFF prose, "what we discussed"
- If parent context contradicts the spec → **disk wins**

## Objection format
- **ID**, **Severity**, **Objection**, **Impact**, **Suggestion**

## Categories to consider
- Ambiguity and untestable acceptance criteria
- Missing edge cases, error paths, empty states
- Security, privacy, abuse, compliance
- Operations: scale, cost, migration, rollback
- Contradictions with `ARCH-000`, existing ADRs, or contracts

## Output

```markdown
## Challenger review — [REQ|ARCH|ADR]-NNN
**Round:** 1 | 2
**Spec path:** ...
**Prior objection IDs considered:** [none | list]

### Objections
1. **ID:** C1 | **Severity:** Blocking|Important|Nit
   - **Objection:** ...
   - **Impact:** ...
   - **Suggestion:** ...

### Non-issues considered
- [category]: why not applicable

### Loop control
- **Approval blocked:** yes | no
- **Blocking open:** [IDs]
- **Recommend:** human-approve-ready | author-revise | human-override-needed | deadlock-human-required
```

End with skill `spec-handoff`.

From sushilti80/specforge-engineering-team, MIT.

More from sushilti80/specforge-engineering-team

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Jul 12, 2026 (3 months ago)
Last release
Aug 6, 2026 (2 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from sushilti80

Cite this page

spec-challenger skill. Agent Mart. https://agentmart.dev/skills/sushilti80-spec-challenger (last updated 2026-10-24)