Wp Abilities Verify

Verify a WordPress plugin’s Abilities API registrations. The centerpiece is the adversarial annotation correctness check: a readonly: true ability that actually writes (via $wpdb->update, update_option, a non-GET delegate, etc.) is a security and UX disaster because agents plan actions on the basis of the annotations they introspect. This skill catches those lies by reading the callback body and comparing what it does against what the annotation claims.

The skill also validates audit docs produced by wp-abilities-audit, checks permission gates and schema hygiene, and optionally executes each ability against a live environment.

When to use

• After abilities have been registered in a plugin but before a PR lands.
• As a health-check on an already-shipped plugin (catch regressions where a refactor turned a readonly ability into a writing one).
• To validate an audit document before handing it to an implementer.

Two modes

• Static mode — runs from the plugin checkout. No env. Enumerates via source inspection, runs the adversarial correctness check, runs schema and permission lints, and validates audit docs.
• Runtime mode — requires a running env. Does everything static does PLUS: wp_get_abilities() for authoritative enumeration, executes each ability with curated inputs, confirms permission roundtrip against real users, and runs a twin-invocation heuristic on idempotent: true abilities to flag candidates for review (return-value equality is a signal, not a verdict — core defines idempotent as “no additional effect on the environment”).

Both modes produce the same structured report format.

A static-mode PASS means “no obvious-shape violations,” not “verified write-free.” For high-stakes plugins, run runtime mode before landing — it catches bootstrap-order, permission-roundtrip, and idempotency issues that static can’t. See references/annotation-correctness.md for the static blind spots.

Inputs required

1. Plugin checkout path — working tree to verify.
2. Mode — static or runtime. Default to static if unspecified.
3. (Runtime only) Env-up command — read the plugin’s AGENTS.md. Common patterns: npm run wp-env start, npx wp-env start, or a composer-based bring-up. Plugin families with their own dev tooling will document their own command. Do NOT assume npm run wp-env works.
4. (Optional) Audit doc path — enables cross-checks between the audit and the registered abilities, and validates the audit itself.
5. Report output path — explicit path, typically the user’s vault.

Prerequisites

• wp-project-triage has been run on the plugin.
• The plugin has at least one registered ability in source. Zero hits on wp_register_ability( → return a clear “no abilities registered” report, not an empty PASS.

Procedure

1. (If audit provided) Validate the audit doc

Read references/audit-schema-validation.md. Validate the audit against the canonical schema owned by wp-abilities-audit. Surface missing required fields, multiple reference_ability: true, and backing: null entries that aren’t paired with a surfaced_gaps entry. backing: null alone is WARN (intentional gap output), not FAIL.

2. Enumerate abilities statically

Read references/static-enumeration.md. Find each wp_register_ability( call, extract the name, the annotation block, and the execute-callback location. Use a multi-line tool (rg --multiline --pcre2) — the canonical formatting splits the call across lines. Record each ability’s source-file + line + annotations + callback byte range.

3. (Runtime only) Enumerate via REST + wp-cli

Read references/runtime-harness.md. Bring the env up using the command from AGENTS.md, then enumerate via wp_get_abilities() over wp-cli and cross-check against the static inventory. Source-only → FAIL (registration not firing). Runtime-only → WARN (dynamic registration path).

4. Annotation correctness (the adversarial core)

Read references/annotation-correctness.md. Read each callback body and verify it matches the annotation claim:

• readonly: true → callback must not write to the database, the options table, post / user / term / comment data, the filesystem, cron, or via non-GET HTTP / REST delegates.
• destructive: false → callback must not delete, refund, void, cancel, or trash.
• idempotent: true → repeated calls with the same input have no additional effect on the environment (per the idempotent annotation’s docblock in class-wp-ability.php). Static catches counter writes and per-call cron schedules; runtime adds a twin-invocation heuristic for visible state changes.

The reference lists common write patterns as a starting set, not a checklist — plugin vocabularies vary, and the agent extends with verbs specific to the plugin under verification.

False positives get suppressed via an inline // verify-ignore: <annotation> -- <reason> comment.

5. Permission roundtrip

Read references/permission-roundtrip.md. Static: classify each permission_callback against the six shapes (preferred Shape A current_user_can(...); FAIL on Shape B-bad WP_REST_Request patterns or Shape E literal true). Runtime: anon and subscriber denied; admin allowed (unless deliberately public). When an audit was provided, cross-check the registered cap against the audit’s declared gate.

6. Schema lints

Read references/schema-lints.md. Six small principles applied to each ability’s input_schema: object schemas declare additionalProperties; required fields have descriptions; enums non-empty; no $ref; defaults are statically constant (including (object) array()); reference abilities have no required inputs.

Cross-reference ../wp-abilities-api/references/input-schema-gotchas.md for the four runtime gotchas (defaults not injected on the property-level path, pagination key drift, empty() on string IDs, direct vs indirect invocation strictness).

7. Error-code vocabulary

Cross-reference ../wp-abilities-api/references/error-code-vocabulary.md. Inspect each callback’s WP_Error returns; non-vocabulary codes → WARN.

Verification

The run produces a structured markdown report at the user-specified path:

---
Last updated: <YYYY-MM-DD HH:MM>
---

# <Plugin> Abilities Verification — <Static|Runtime> Mode

## Status: <PASS|WARN|FAIL>

## Audit doc validation (if provided)

## Static inventory

## Annotation correctness
| Ability | Claim | Result | Evidence |
|---|---|---|---|

## Permission gates

## Schema lints

## Error-code vocabulary

Every ability is OK, WARN, or FAIL. A single FAIL → top-line FAIL; WARNs without FAILs → WARN; otherwise PASS.

Failure modes / debugging

• Env not reachable (runtime) — env-up failed or Docker isn’t running. Re-run wp-project-triage, then fix the env. Don’t fall back silently to static without noting it in the report.
• No abilities in source — return a clear “nothing to verify” report.
• Audit schema mismatch — point at references/audit-schema-validation.md; don’t auto-fix the audit.
• False positive on readonly-writes — see the // verify-ignore mechanism in references/annotation-correctness.md. Document why each suppression is legitimate.
• Runtime enumeration smaller than static — registration hook isn’t firing. Check init hook timing, activation state, autoloader order.

Escalation

• Recurring legitimate pattern that trips the adversarial check across multiple plugins → propose adding it to the suppression guidance in annotation-correctness.md. Don’t broaden the candidate-pattern list speculatively.
• Audit-schema validator rejects a legitimate audit → the canonical schema in ../wp-abilities-audit/references/audit-schema.md has evolved. Update references/audit-schema-validation.md to match.

Out of scope

Token-budget measurement is a separate verification axis — an annotation-clean, schema-clean, runtime-passing ability set can still be unshippable if its tools/list form burns through an agent’s context budget. That axis is tracked separately. Do not aggregate manual or external measurement into this skill’s PASS / FAIL verdict.