scoped-differential-review
Wraps a vendored differential-review skill and routes its findings against a GitHub issue-label universe.
Skill by kyzooghost
Chalk Labs
Need help deploying scoped-differential-review?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
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/kyzooghost-scoped-differential-review
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.
About
Use cases: security
Reviews
Write a reviewNo 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/scoped-differential-review
What's inside
- Skill name
scoped-differential-review- Description
- Wraps a vendored differential-review skill and routes its findings against a GitHub issue-label universe. Classifies each finding as IN SCOPE, OWNED BY ANOTHER TICKET, or GENUINE GAP. Proposes gap tickets for unowned findings. Use when running a scope-bounded security review of a PR.
- License
- MIT
- Instructions
- 84 lines in SKILL.md · view on GitHub
Files (15)
- ATTRIBUTION.md
- SKILL.md
- gap-ticket-template.md
- scope-routing.md
- LICENSE · vendor/differential-review/
- README.md · vendor/differential-review/
- adversarial-modeler.md · vendor/differential-review/agents/
- diff-review.md · vendor/differential-review/commands/
- SKILL.md · vendor/differential-review/skills/differential-review/
- adversarial.md · vendor/differential-review/skills/differential-review/
- openai.yaml · vendor/differential-review/skills/differential-review/agents/
- trail-of-bits-mark.svg · vendor/differential-review/skills/differential-review/assets/
- methodology.md · vendor/differential-review/skills/differential-review/
- patterns.md · vendor/differential-review/skills/differential-review/
- reporting.md · vendor/differential-review/skills/differential-review/
SKILL.md (first lines)
# Scoped Differential Review
Runs a vendored security review skill on a PR, then classifies every finding against a scope map built from a GitHub issue-label universe. Findings route to target tickets, defer to other tickets, or become proposed gap tickets.
## Prerequisites
- `gh` CLI authenticated with read/write access to the target repository's issues.
## Inputs
- `PR` - URL or `owner/repo#N` of the pull request under review.
- `--tickets` - one or more target ticket numbers this PR implements (comma-separated, e.g. `42,43`). Required.
- `--label` - the issue label defining the ticket universe (e.g. `my-feature`). Required.
- `--repo` - the repository containing the labeled issues. Inferred from `PR` when possible; required otherwise.
## Workflow
Five phases. Phases 1-2 prepare the scope map, phase 3 delegates to the vendored review skill, phases 4-5 route findings and handle gaps.
### Phase 1 - Parse inputs and build scope map
1. Parse `PR`, `--tickets`, `--label`, `--repo`.
2. Fetch all open tickets in the label universe:
gh issue list -R <REPO> --label <LABEL> --state open --limit 200 --json number,title,body
3. For every ticket, comprehend the full body and extract what it OWNS and what it EXCLUDES. Use any available signal: headings, acceptance criteria, task lists, descriptions, "out of scope" language, title. See [scope-routing.md](scope-routing.md) for the full extraction and classification rules.
4. Build the scope map: ticket number -> OWNS / EXCLUDES.
5. If a ticket is too vague to extract any ownership, mark it "unparseable" in the scope map. Continue with the rest.
### Phase 2 - Validate target tickets
1. For each ticket in `--tickets`, fetch via `gh issue view <N> -R <REPO>`.
2. Confirm it is OPEN and carries `--label`. Stop if any target is closed, missing, or unlabeled.
3. Confirm at least one target ticket has a parseable ownership boundary. If all are unparseable, stop and report the error.
### Phase 3 - Run differential-review
1. Read and follow `vendor/differential-review/skills/differential-review/SKILL.md`. The vendored skill runs its own methodology unmodified and produces a findings report.
2. The adversarial modeler agent definition is at `vendor/differential-review/agents/adversarial-modeler.md`. Reference it for high-risk findings as the vendor skill directs.
3. Do not alter the vendor skill's methodology, risk classification, or report structure.
### Phase 4 - Route findings by scope
For every finding in the vendor's report, apply the A/B/C classification from [scope-routing.md](scope-routing.md):
1. Map the finding to a scope-map entry. Cite the ticket number and ownership statement used.
2. Classify as (A) IN SCOPE, (B) OWNED BY ANOTHER TICKET, or (C) GENUINE GAP.
3. Stubs and TODOs standing in for other tickets' work are correct by design. Verify the stub matches the interface; do not raise as findings.
### Phase 5 - Emit report and handle gaps
1. Emit the vendor's full report unchanged, including its verdict.
2. Append a `## Scope Routing` section containing:
- **Scope Map** - the extracted scope map for auditability.
- **Scope-adjusted verdict** - computed per the reconciliation rules in [scope-routing.md](scope-routing.md).
- **In-scope findings (A)** - each with target-ticket assignment, severity, file/line.
- **Deferred - owned by other tickets (B)** - one-liners with ticket numbers.
- **Scope violations** - PR work belonging to other tickets or no ticket.
- **Proposed new tickets (C)** - full drafted ticket bodies using [gap-ticket-template.md](gap-ticket-template.md), or "No gaps found."
3. If gap tickets are proposed, request one batch approval. On approval, file each via `gh issue create -R <REPO> --label <LABEL> -t "<title>" -b "<body>"`, capture real numbers, substitute back into the report. On decline, leave drafts unfiled.
## Invocation
```
/scoped-differential-review <PR-URL> --tickets 42,43 --label my-feature --repo org/project
```
`--repo` is optional when inferable from the PR URL. All scope-routing guarantees (no-duplication, no-new-scope, gap routing) are built-in defaults.
## Error handling
- `--tickets` omitted -> stop, ask for target tickets.
- Target ticket not in label universe -> stop, report which ticket, ask to confirm.
- Empty label universe (zero open tickets with `--label`) -> stop, ask to confirm the label.
- All target tickets unparseable -> stop, report the error.
- `differential-review` errors or emits no report -> stop, surface the error.
- `gh issue create` fails mid-batch -> report error, stop. Already-filed tickets remain; report their numbers.
## Mutations
The skill's only GitHub mutation is `gh issue create` for approved gap tickets. It never posts PR comments, edits existing issues, changes labels, or closes issues.From kyzooghost/skills, MIT.
More from kyzooghost/skills
- 0.570score
pr-review-comments
kyzooghost
Transform security review findings into targeted GitHub PR review comments.
SkillAFreshness: AClaude Code +9 - 0.538score
fix-ci
kyzooghost
Use when the user asks for /fix-ci behavior or the equivalent command workflow.
SkillAFreshness: AClaude Code +9 - 0.529score
ship-from-plan
kyzooghost
Use when the user provides a completed implementation plan and wants one continuous isolated execution through a reviewed, CI-green pull request marked ready…
SkillBFreshness: BClaude Code +9 - 0.529score
create-scoped-tickets
kyzooghost
Decompose a technical implementation spec into mutually exclusive GitHub Agent Work Tickets under one or more labels, either seeding an empty label universe…
SkillBFreshness: BClaude Code +9 - 0.524score
scoped-planning
kyzooghost
Per-stage scope-check overlay across the planning lifecycle.
SkillBFreshness: BClaude Code +9 - 0.524score
complexity-review
kyzooghost
Review a PR for unnecessary complexity, brittle test strategies, and overengineering.
SkillBFreshness: BClaude Code +9 - 0.479score
discover-repo
kyzooghost
Generate a comprehensive AGENTS.md index for a repository with a CLAUDE.md symlink.
SkillCFreshness: CClaude Code +9 - 0.479score
sync-command
kyzooghost
Sync a command from this repo into local agent surfaces.
SkillCFreshness: CClaude Code +9 - 0.479score
sync-skill
kyzooghost
Sync a skill from this repo into the local machine's skill directories.
SkillCFreshness: CClaude Code +9 - 0.479score
llm-wiki
kyzooghost
Build and maintain a personal knowledge wiki using LLMs.
SkillCFreshness: CClaude Code +9 - 0.474score
perf-review
kyzooghost
Use when drafting a periodic performance review (annual review, mid-year, promo packet, peer feedback) from monthly win reports and supporting sources, where…
SkillCFreshness: CClaude Code +9 - 0.474score
handover
kyzooghost
Prepare a structured handover document from the current session for a fresh Claude session to continue the work.
SkillCFreshness: CClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Aug 13, 2026 (2 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/kyzooghost/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from kyzooghost
- 0.570score
pr-review-comments
kyzooghost
Transform security review findings into targeted GitHub PR review comments.
SkillAFreshness: AClaude Code +9 - 0.538score
fix-ci
kyzooghost
Use when the user asks for /fix-ci behavior or the equivalent command workflow.
SkillAFreshness: AClaude Code +9 - 0.529score
create-scoped-tickets
kyzooghost
Decompose a technical implementation spec into mutually exclusive GitHub Agent Work Tickets under one or more labels, either seeding an empty label universe…
SkillBFreshness: BClaude Code +9 - 0.529score
ship-from-plan
kyzooghost
Use when the user provides a completed implementation plan and wants one continuous isolated execution through a reviewed, CI-green pull request marked ready…
SkillBFreshness: BClaude Code +9 - 0.524score
scoped-planning
kyzooghost
Per-stage scope-check overlay across the planning lifecycle.
SkillBFreshness: BClaude Code +9 - 0.524score
complexity-review
kyzooghost
Review a PR for unnecessary complexity, brittle test strategies, and overengineering.
SkillBFreshness: BClaude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get scoped-differential-review running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
scoped-differential-review skill. Agent Mart. https://agentmart.dev/skills/kyzooghost-scoped-differential-review (last updated 2026-10-24)