Agent skill

Ad Review

by CorridorTech in CorridorTech/PoseCap

Two-axis fresh-context code review per WORKFLOW §10. An agent skill from CorridorTech/PoseCap.

Apache-2.0Auto-check: notesDevelopment

Install Ad Review

skills CLI
$ npx skills add CorridorTech/PoseCap --skill ad-review -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install CorridorTech/PoseCap ad-review --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/CorridorTech/PoseCap.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/ad-review .claude/skills/ad-review && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
ad-review
GitHub stars
224
Token cost
~2.4k tokens
SKILL.md length
994 words
Files
1
Skills in repo
15
Repo updated
First seen
Licence
Apache-2.0

At a glance

Two-axis fresh-context code review per WORKFLOW §10. An agent skill from CorridorTech/PoseCap.

  • Works in 7 steps: Scope the review → Identify Standards sources → Identify the Spec source → …
  • The user wants to review a diff
  • SKILL.md covers Step 0 — Scope the review, Step 1 — Identify Standards…, Step 2 — Identify the Spec… and Step 3 — Build two…, plus 5 more sections
  • Calls git and gh

What it does

Ad Review is an agent skill from CorridorTech/PoseCap. Two-axis fresh-context code review per WORKFLOW §10. Splits the review into Standards (does the diff conform to AGENTS.md / ARCHITECTURE.md / GUIDELINES.md / CONTEXT.md / accepted ADRs?) and Spec (does the diff match the originating task / spec / PRD?), runs them as parallel sub-agent passes so neither axis masks the other, then aggregates findings side-by-side. Use when the user wants to review a diff, branch, PR, or recent commits against the project's spec, audit for bugs / coupling / edge cases / spec drift…

Its SKILL.md is about 2.4k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Development, covering Subagents, Code review and Architecture decision records. It works with Git. The licence is Apache-2.0.

When your agent uses it

  • The user wants to review a diff
  • Recent commits against the projects spec
  • Audit for bugs / coupling / edge cases / spec drift
  • Run a §10 senior-reviewing-junior pass

Example prompts

  • “approve”
  • “/ad-review”

Requirements

  • Pre-approved tools (allowed-tools): Read, Glob, Grep, Bash, Task

Workflow steps

7 steps, taken from the step headings in SKILL.md.

  1. Scope the review
  2. Identify Standards sources
  3. Identify the Spec source
  4. Build two axis-bounded handoffs
  5. Persist both handoffs to disk
  6. Spawn both sub-agents in parallel
  7. Aggregate

What it can do on your machine

Read from SKILL.md and the folder at commit 626701b. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves these tools, so the agent can use them without asking each time:

    • Read
    • Glob
    • Grep
    • Bash
    • Task

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • git
    • gh

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    Links to these hosts (documentation or services it may open):

    • github.com

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Ad Review loads about 2.4k tokens when it runs. Until then it costs about 157 tokens; SKILL.md has 994 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~157
When it runs · the whole SKILL.md, loaded when a task matches
~2.4k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check: notes

The automated check noted patterns worth knowing about, such as sudo or a known installer.

  • NotePre-approves every shell command (allowed-tools: Bash)SKILL.md
    allowed-tools: Read, Glob, Grep, Bash, Task

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from CorridorTech/PoseCap at commit 626701b, republished under its Apache-2.0 licence (© CorridorTech). 994 words, ~2,401 tokens.

Download SKILL.mdSave it as .claude/skills/ad-review/SKILL.md (or your agent's skills folder).
name
ad-review
description
Two-axis fresh-context code review per WORKFLOW §10. Splits the review into Standards (does the diff conform to AGENTS.md / ARCHITECTURE.md / GUIDELINES.md / CONTEXT.md / accepted ADRs?) and Spec (does the diff match the originating task / spec / PRD?), runs them as parallel sub-agent passes so neither axis masks the other, then aggregates findings side-by-side. Use when the user wants to review a diff, branch, PR, or recent commits against the project's spec, audit for bugs / coupling / edge cases / spec drift, or run a §10 senior-reviewing-junior pass. Adversarial framing — never emits an "approve" verdict.
allowed-tools
Read, Glob, Grep, Bash, Task
summary
Two-axis code review per WORKFLOW §10. Claude Code uses fresh-context subagents; Codex writes an audit trail, reviews inline by default, and ships a reviewer…

/ad-review

Implements WORKFLOW §10 (Reviewer With Fresh Context). The current session is biased about the code it produced — the same reasoning that wrote it defends it. This skill assembles two clean handoffs (Standards-axis, Spec-axis), delegates each to a sub-agent that starts with no history, and aggregates the findings side-by-side. The two axes are deliberately separate so a Spec pass cannot mask a Standards fail (and vice versa) — the dichotomy is borrowed from mattpocock/skills/review and bound to this kit's six-layer artifact stack.

Step 0 — Scope the review

Confirm what to review. Default scopes, in priority order:

  1. User-named ref or PR (/ad-review main..HEAD, /ad-review PR#42, /ad-review <commit-sha>).
  2. Current branch vs main (git diff main...HEAD).
  3. Working-tree changes (git diff plus git diff --staged).

If no diff exists, stop and tell the user — there's nothing to review.

When the host exposes AskUserQuestion, use it at Step 0 to confirm the review scope as a multi-choice card (branch vs main / PR#NN / commit-sha / working-tree) instead of asking inline text. Falls back to numbered text on hosts without the primitive (Codex).

Capture the diff command once: git diff <range> (use ... three-dot for ref-vs-ref so the comparison is against the merge-base). Note the commit list with git log <range> --format=%B.

Handoff-integrity gate. Compute the commit count: git rev-list --count <range>. Bind it to N. The handoff header must include the literal line Range: <range> (N commits). The ## Spec slice — commit messages section must contain exactly N ### <sha> <subject> entries. If your body has fewer, you mis-bounded the range — stop and re-scope. The subagent's review signal is bounded above by handoff fidelity; a quiet count mismatch produces silently-incomplete reviews.

Step 1 — Identify Standards sources

Anything in the repo that documents how code should be written. Read what exists; do not fabricate references.

  • AGENTS.md at the repo root.
  • ARCHITECTURE.md at the repo root.
  • GUIDELINES.md at the repo root.
  • CONTEXT.md at the repo root, or CONTEXT-MAP.md plus per-context CONTEXT.mds.
  • Every ADR under doc/adr/ with Status: accepted whose subject is touched by the diff. When in doubt, include rather than skip.
  • CONTRIBUTING.md if present.
  • Machine-enforced standards (.editorconfig, eslint.config.*, biome.json, prettier.config.*, tsconfig.json) — note their presence, but instruct the Standards sub-agent to skip what tooling already enforces.

Step 2 — Identify the Spec source

In this order, take the first that resolves:

  1. Task references in the diff or recent commit messages (Task NNNN, 0NNN-, Closes task-0042) → read the file's Acceptance Criteria and Plan sections.
  2. An originating spec under doc/specs/ whose filename matches the branch name or the dominant feature touched by the diff.
  3. A parent PRD under doc/product/ referenced by the spec.
  4. Issue references in commit messages (#123, Closes #45) — fetch via gh issue view if the kit has gh available.

If nothing resolves, mark the Spec axis as no spec source provided and the Spec sub-agent will skip with a one-line note rather than fabricate findings.

Step 3 — Build two axis-bounded handoffs

Each sub-agent will receive only the slice for its axis. Do not paste the Spec slice into the Standards handoff or vice versa — that's the bias the split exists to prevent.

Standards handoff (<scope>-standards.md):

=== AGENTIC-REVIEW HANDOFF — STANDARDS AXIS ===

Axis: Standards. Report only findings that violate documented standards or
introduce bugs / coupling / edge-case gaps. Skip what tooling enforces.
Skip Spec-axis findings (missing requirements, scope creep) — a separate
sub-agent covers those.

--- DIFF ---
<git diff output>

--- STANDARDS SOURCES ---
<AGENTS.md, ARCHITECTURE.md, GUIDELINES.md, CONTEXT.md / CONTEXT-MAP.md,
applicable accepted ADRs (full text), CONTRIBUTING.md if present>

--- TOOLING NOTE ---
<list of machine-enforced configs found — eslint.config.*, biome.json,
.editorconfig, tsconfig.json, etc. The sub-agent must NOT re-check what
tooling already enforces.>

=== END HANDOFF ===

Spec handoff (<scope>-spec.md):

=== AGENTIC-REVIEW HANDOFF — SPEC AXIS ===

Axis: Spec. Report only:
  (a) requirements the spec asked for that are missing or partial;
  (b) behaviour in the diff that wasn't asked for (scope creep);
  (c) requirements that look implemented but where the implementation
      looks wrong against the spec line.
Quote the spec line for each finding. Skip Standards-axis findings — a
separate sub-agent covers those.

--- DIFF ---
<git diff output>

--- SPEC SOURCES ---
<task file Acceptance Criteria + Plan, originating spec (full text),
parent PRD (full text), recent commit messages for the range, originating
issue body if fetched>

=== END HANDOFF ===

If Step 2 found no spec, write the Spec handoff as a single block: no spec source provided — report exactly that and stop.

Show full SKILL.md (449 more words)Show less

Step 4 — Persist both handoffs to disk

Write both files to .agentic/reviews/<ISO-timestamp>-<scope-slug>-{standards,spec}.md at the repo root. Create the directory if it does not exist. The <scope-slug> encodes the review target (branch-vs-main, pr-42, commit-abc1234, working-tree).

After write, print wc -l <path> alongside each handoff path so the user has a sanity-check value. Verify the handoff-integrity gate from Step 0 still holds: the commit-message section must have exactly N ### <sha> entries where N = git rev-list --count <range>. If it doesn't, do not dispatch — re-scope and rebuild.

Advise the user to add .agentic/reviews/ to their .gitignore if it is not already — handoffs are ephemeral per-review artifacts, not committed history.

If the combined diff spans >50 files, ask the user to narrow scope before invoking the sub-agents — the prompt cost compounds across two passes.

Step 5 — Spawn both sub-agents in parallel

Send a single message with two Task tool calls, both routing to the bundled fresh-context-reviewer subagent. Each call receives its own axis-specific handoff as the prompt. Parallel dispatch is critical — sequential would cost 2× wall time for no rigor gain.

If the Spec axis was marked no spec source provided, dispatch only the Standards Task call. Note the skipped axis in the final report.

Step 6 — Aggregate

Present the two sub-agents' reports verbatim, under explicit headings, in this order:

## Standards Findings

<verbatim output from the Standards sub-agent>

## Spec Findings

<verbatim output from the Spec sub-agent, or "Spec: skipped — no spec source provided">

Do not merge or rerank findings — the two axes are deliberately separate so the user can see them independently. Do not synthesize an overall "approve" verdict.

End with a one-line aggregate summary:

Aggregate: <N Standards Blockers, M Standards Concerns> / <P Spec Blockers, Q Spec Concerns>. Worst: <one-line quote of the highest-severity finding from either axis>.

Reference both persisted handoff paths in your reply so the user can audit what was sent to each axis.

Output contract

  • Two persisted handoff files at .agentic/reviews/<ISO>-<scope>-standards.md and .agentic/reviews/<ISO>-<scope>-spec.md (the latter may be a single-line "no spec source provided" stub).
  • Two parallel Task invocations of fresh-context-reviewer, each with its axis-bounded handoff (or one invocation if Spec was skipped).
  • Aggregated reply under ## Standards Findings and ## Spec Findings headings, verbatim, no cross-axis re-ranking.
  • One-line aggregate summary at the end with counts per axis and the worst single finding.
  • Both persisted handoff paths cited.
  • No "approve" verdict, no defending of the code, no rewrite of the diff. Empty axis result is reported explicitly.

Next

  • Address every Standards Blocker before merge — that's the code-quality hard gate. Re-run /ad-review on the fix to confirm it cleared.
  • Address Spec Blockers next — implementation-vs-spec drift is the second hard gate.
  • Each Concern (from either axis) becomes a follow-up /ad-task; do not let them silently accumulate.
  • Notes are informational; close them out in the original task's Notes log if relevant.
  • If the Spec axis was skipped, decide whether a /ad-spec is overdue — work without a spec means future reviews are Standards-only.
  • Once both axes are clear: merge per project conventions.

© CorridorTech, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .claude/skills/ad-review of CorridorTech/PoseCap.

Open the folder on GitHubat commit 626701b

Compare with similar skills

Ad Review next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Ad Review compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Ad Review this skillCorridorTech/PoseCap224—~2.4kAutomated safety check: NotesApache-2.0
CCPM Project Managementautomazeio/ccpm8.4k—~1.1kAutomated safety check: PassMIT
Cursor Composer Task DelegateChachamaru127/claude-code-harness3.2k—~4.4kAutomated safety check: NotesMIT
Setup Matt Pocock Skillsywwynm/EverythingDone1449 repos~1.7kAutomated safety check: PassGPL-3.0
Reviewfossasia/eventyay-interpretation1.6k35 repos~996Automated safety check: PassApache-2.0
Cxas Agent FoundryGoogleCloudPlatform/cxas-scrapi107—~2.4kAutomated safety check: PassApache-2.0

Similar skills

  • Runs a spec-driven workflow from PRD to epic to GitHub issues to parallel agents, with status, standup and blocked-work reports from bundled scripts.

    8.4k GitHub stars~1.1k tokensUpdated 6 mo ago
    Product & Project ManagementAuto-check passed
  • Cursor Composer Task Delegate

    Chachamaru127/claude-code-harness

    Hands one implementation task to Cursor Composer in an isolated git worktree, then reviews its diff and cherry-picks the result into the main branch.

    3.2k GitHub stars~4.4k tokensUpdated 3 days ago
    DevelopmentAuto-check: notes
  • Setup Matt Pocock Skills

    ywwynm/EverythingDone

    Sets up an Agent skills block in AGENTS.md/CLAUDE.md and docs/agents/ so the engineering skills know this repo's issue tracker (GitHub or local markdown), triage label vocabulary, and domain doc…

    144 GitHub starsUsed in 9 repos~1.7k tokens
    Agent WorkflowsAuto-check passed
  • Review

    fossasia/eventyay-interpretation

    Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes — Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match…

    1.6k GitHub starsUsed in 35 repos~996 tokens
    Product & Project ManagementAuto-check passed
  • Cxas Agent Foundry

    GoogleCloudPlatform/cxas-scrapi

    End-to-end GECX/CXAS/CES conversational agent lifecycle -- build agents from requirements (PRD-to-agent), create and run evals (goldens, simulations, tool tests, callback tests), debug failures, and…

    107 GitHub stars~2.4k tokensUpdated today
    DevelopmentAuto-check passed
  • Workflow Code Review

    davidYichengWei/agentic-engineering-framework

    代码评审。协调 5 个专项 reviewer subagent 对代码进行并行多维度审查。可由用户直接触发,也可由主 agent 加载后作为 Judge 执行。

    158 GitHub stars~1.1k tokensUpdated 6 mo ago
    DevelopmentAuto-check passed

More from CorridorTech/PoseCap

All 15 skills in this repo
  • Ad Adr

    CorridorTech/PoseCap

    Draft a new ADR (Architecture Decision Record) at doc/adr/NNNN-<short-title.md, using Michael Nygard's Context/Decision/Consequences/Alternatives pattern.

    224 GitHub stars~1k tokensUpdated yesterday
    Auto-check: notes
  • Ad Architecture

    CorridorTech/PoseCap

    Generate ARCHITECTURE.md at the repo root by scanning the code first, pre-filling layers/patterns/observability/deployment from observed signals, then asking only the genuine gaps.

    224 GitHub stars~1.5k tokensUpdated yesterday
    Auto-check: notes
  • Ad Archive

    CorridorTech/PoseCap

    Hard-delete completed plan files (tasks Status:done, specs Status:shipped, PRDs Status:superseded, ADRs Status:superseded or deprecated) via git rm, leaving git history as the only ledger.

    224 GitHub stars~1.9k tokensUpdated yesterday
    Auto-check: notes
  • Ad Drift

    CorridorTech/PoseCap

    Read-only drift audit — compare AGENTS.md, ARCHITECTURE.md, ADR statuses, feature specs in doc/specs/, and documentation discipline against what the code actually does.

    224 GitHub stars~1.8k tokensUpdated yesterday
    Auto-check: notes
  • Ad Hooks

    CorridorTech/PoseCap

    Scaffold deterministic quality gates per WORKFLOW.md §11 — pre-commit (lint, format, secret-scan), pre-push (build, unit, integration).

    224 GitHub stars~2.4k tokensUpdated yesterday
    Auto-check: notes
  • Ad Next

    CorridorTech/PoseCap

    Survey the project's state across the six-layer artifact stack and recommend prioritized next actions, modeled on flutter doctor.

    224 GitHub stars~3.3k tokensUpdated yesterday
    Auto-check: notes

Works with

Questions about Ad Review

What does Ad Review do?

Two-axis fresh-context code review per WORKFLOW §10. An agent skill from CorridorTech/PoseCap. Ad Review is an agent skill from CorridorTech/PoseCap. Two-axis fresh-context code review per WORKFLOW §10.

When should I use Ad Review?

Ad Review fits situations like: the user wants to review a diff; recent commits against the projects spec; audit for bugs / coupling / edge cases / spec drift; run a §10 senior-reviewing-junior pass.

How do I install Ad Review in Claude Code?

Run `npx skills add CorridorTech/PoseCap --skill ad-review -a claude-code`. Or copy the skill folder (.claude/skills/ad-review in CorridorTech/PoseCap) into .claude/skills/ad-review in your project. Claude Code loads it when a task matches its description.

How do I install Ad Review in Codex?

Run `npx skills add CorridorTech/PoseCap --skill ad-review -a codex`. Or copy the skill folder (.claude/skills/ad-review in CorridorTech/PoseCap) into .agents/skills/ad-review in your project. Codex loads it when a task matches its description.

Can I use Ad Review in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add CorridorTech/PoseCap --skill ad-review -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/ad-review, .gemini/skills/ad-review, .github/skills/ad-review and .opencode/skills/ad-review in your project.

What does Ad Review need to run?

Going by SKILL.md and its folder, Ad Review needs the command-line tools its instructions call (git and gh). Its frontmatter pre-approves these tools: Read, Glob, Grep, Bash, Task.

Does Ad Review access the network?

SKILL.md names 1 domain. As links in the text: github.com. This is read from the text; nothing was executed.

Is Ad Review safe to install?

Our automated static check of SKILL.md found notes only (pre-approves every shell command (allowed-tools: bash)), nothing it rates as a warning. It is not a guarantee. Review the folder before installing.

What licence does Ad Review use?

Ad Review is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Ad Review use?

About 2.4k tokens (SKILL.md is roughly 9.6k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Ad Review?

Skills that share tags, products or a category with Ad Review: CCPM Project Management (automazeio/ccpm, 8.4k stars), Cursor Composer Task Delegate (Chachamaru127/claude-code-harness, 3.2k stars), Setup Matt Pocock Skills (ywwynm/EverythingDone, 144 stars) and Review (fossasia/eventyay-interpretation, 1.6k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Ad Review?

CorridorTech (a GitHub organization) maintains it in CorridorTech/PoseCap, which has 224 GitHub stars. The repository holds 15 skills in this directory. The repository was last updated on October 7, 2026.

Source: CorridorTech/PoseCap on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.