AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Review Api Compat

skill-paultyng-skill-issue-review-api-compat · by paultyng

Use when the user asks for an API compatibility review, breaking change review, proto breaking change review, buf breaking review, OpenAPI compatibility check, gRPC backwards compatibility audit, "are these API changes breaking", "did we break the wire", contract evolution review, or backwards-compatibility audit on an API surface.

No reviews yet
0 installs
33 views
0.0% view→install

Install

$ agentstack add skill-paultyng-skill-issue-review-api-compat

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-paultyng-skill-issue-review-api-compat)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
2mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Review Api Compat? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

API Compatibility Review

Structured review of API contract evolution. Detects breaking changes in protobuf services, OpenAPI specs, and (where applicable) GraphQL schemas; classifies them as wire-breaking, behavior-breaking, or policy-breaking; recommends remediation (deprecate-then-remove, version bump, new field).

Out of scope (defer to siblings):

  • Internal Go API surface (function signatures, struct fields) — review-code covers this through SOLID/coupling lens
  • New field validation correctness — review-code / review-security
  • Protobuf style / lint pass-fail (naming, comments, package layout) — review-code runs buf lint

review-api-compat is the diff-aware counterpart: same proto/OpenAPI files, but compared against a base ref to catch what changed.

Workflow

1. Scope and explore

  • Confirm scope with the user: changed files (PR or branch diff — the default), explicit paths to a previous version (e.g. a tag), or full codebase against main.
  • Resolve scope. This skill is fundamentally diff-based:
  • Default: base_ref is main (or the PR base). Compute changed files via git diff --name-only --diff-filter=d ...HEAD. Filter to API spec files (see below).
  • PR: use gh pr view --json baseRefName -q .baseRefName for the base, gh pr diff --name-only for changed files.
  • Tag-vs-tag: user supplies ...; same diff machinery.
  • If invoked from review-all: receive file_list, has_changes, base_ref, REVIEW_DIR, and pr_url. If has_changes is false, exit early with status "no API spec changes in scope" — there is nothing to compare.

File classification:

  • Protobuf: *.proto files; presence of buf.yaml / buf.gen.yaml / buf.lock confirms a buf workspace.
  • OpenAPI: openapi.yml, openapi.yaml, openapi.json, swagger.yml, swagger.yaml, swagger.json, files with top-level openapi: or swagger: keys.
  • GraphQL (lighter coverage): *.graphql, *.gql, files with schema { ... } or type Query { ... }.

2. Determine compatibility policy

Detect the project's stance from configuration; ask the user if unclear:

  • Protobuf: buf.yaml breaking: block (rule set: FILE, PACKAGE, WIRE, WIRE_JSON); breaking.except: exclusions; buf.yaml modules with versioning (v1, v1beta1, v2).
  • OpenAPI: SemVer in info.version; declared compatibility policy in repo docs (e.g. API.md); existing release notes for prior breaks.
  • GraphQL: deprecation policy; usage of @deprecated directive.

Record the resolved policy. Findings are graded against it (a WIRE policy allows JSON-but-not-wire breaks; a WIRE_JSON policy doesn't).

3. Run the breaking-change tools

# Protobuf — against the base ref (preferred, integrates with buf.yaml policy)
command -v buf >/dev/null && buf breaking --against "$(git rev-parse $base_ref).git#format=git"

# Protobuf — alternative when buf is not configured
command -v protolock >/dev/null && protolock status --lockdir . --uptodate

# OpenAPI — multiple options; pick whichever is on PATH
command -v oasdiff >/dev/null && oasdiff breaking  
command -v openapi-diff >/dev/null && openapi-diff  

# GraphQL
command -v graphql-inspector >/dev/null && graphql-inspector diff  

To get the base version of a spec file: git show $base_ref: into a temp file, then diff.

Tool output is input to the review; the subagent triages and contextualizes.

4. Launch investigation subagent

Launch a single investigation subagent (subagent_type="generalPurpose", model: sonnet per subagent-model-routing) with: resolved policy, tool outputs, and the changed-spec file list.

Prompt it to:

  • For each detected change, classify as: wire-breaking / behavior-breaking / policy-breaking / non-breaking. See [reference.md](reference.md) for the matrix.
  • For each breaking change, recommend remediation: deprecate-then-remove, version bump, new alternative field/method, additive change instead.
  • For each non-breaking change, sanity-check that it's truly additive (no enum value reuse, no field number reuse, no required→optional flip in a strict consumer).
  • Cite the wire/behavior rule violated.
  • Search the repo for any documented intent (CHANGELOG.md, MIGRATION.md, doc comments) that legitimizes a deliberate break.

5. Present results

Resolve the review output directory (same pattern as siblings):

REVIEW_DATE=$(date +%Y-%m-%d)
REVIEW_DIR=".reviews/${REVIEW_DATE}"
if [ -d "$REVIEW_DIR" ]; then REVIEW_DIR=".reviews/${REVIEW_DATE}-$(date +%H%M)"; fi
~/.claude/scripts/ensure-gitignore.sh '.reviews/'
mkdir -p "$REVIEW_DIR"

Capture run metadata (see [Run metadata header](#run-metadata-header)) and prepend to ${REVIEW_DIR}/API-COMPAT-REVIEW.md.

Output structure:

  1. Run metadata header
  2. Resolved policy (which rule set + exclusions)
  3. Findings table (grouped by surface: proto / openapi / graphql)
  4. Tool availability notes
  5. Recommended remediation order

Present the report to the user.


Run metadata header

RUN_DATETIME=$(date -u +"%Y-%m-%d %H:%M UTC")
GIT_BRANCH=$(git rev-parse --abbrev-ref HEAD)
GIT_COMMIT=$(git rev-parse --short HEAD)
GIT_COMMIT_FULL=$(git rev-parse HEAD)
GIT_SUBJECT=$(git log -1 --pretty=%s)
BASE_REF=; BASE_COMMIT=$(git rev-parse --short "$BASE_REF")
> **Run:** {RUN_DATETIME}
> **Branch:** {GIT_BRANCH} @ {GIT_COMMIT} (`{GIT_COMMIT_FULL}`)
> **Subject:** {GIT_SUBJECT}
> **Base:** {BASE_REF} @ {BASE_COMMIT}
> **Scope:** {scope description}
> **Policy:** {resolved policy, e.g. buf FILE + WIRE}

Finding link wrapping (PR mode)

When pr_url is provided (or gh pr view --json url -q .url 2>/dev/null returns one for standalone runs), wrap every path:line reference inside the finding tables as a Markdown link:

~/.claude/scripts/pr-deeplink.sh "$pr_url"  

The display text stays path:line. Pass L for findings about removed code. Findings follow terse-comments: concrete fix, optional bug:/risk:/nit:/unsure: prefix.


Output Templates

API compatibility findings

| Priority | Surface | Change | Class | Recommendation | Tracked |
|----------|---------|--------|-------|----------------|---------|
| P0 | proto | `pkg.Service.Method` removed at file:line | wire-breaking | Deprecate first; remove in next major version | — |
| P1 | openapi | `GET /users` response field `name` type changed | behavior-breaking | Add new field, keep old; remove in next major | #123 |
| P2 | proto | New field `created_at` added | non-breaking | OK; document in changelog | — |

Class column values: wire-breaking, behavior-breaking, policy-breaking, non-breaking.

Re-evaluation table (for follow-up reviews)

| Finding | Status | What Changed |
|---------|--------|--------------|
| ~~1. Description~~ | RESOLVED | Reverted; field re-added |
| 2. Description | Still applicable | No changes |

Guidelines

  • The default position on a wire-breaking change is block unless an explicit version bump or migration plan is in the PR.
  • A non-breaking classification still warrants a changelog entry if the change is user-visible.
  • Search the organization's codebase (Sourcegraph, GitHub) for consumers of the changed surface before scoring a break as low-impact. A removed proto field that nobody reads is annoying; one that 200 services read is a P0.
  • When the user asks for a follow-up review, find the most recent review directory and append the re-evaluation table.
  • For detailed framework categories, see [reference.md](reference.md).
  • REVIEW.md integration: If a REVIEW.md context section was provided by the review-all orchestrator (or exists at the repository root when running standalone), treat its rules as additional review criteria. "Always check" items are HIGH severity; domain-specific items (API compatibility section) are MEDIUM severity. "Skip" patterns exclude matching files from review scope.
  • Findings must cite probed evidence (path:line, grep output, command result), not pattern-matched suspicion. Per ~/.claude/rules/probe-not-assume.md.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.