How do I install the Verify skill?

Run npx skillmds@latest add nvzver/verify in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Verify safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Verify?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Verify free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Verify?

NVZver (@nvzver) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Verify by nvzver — Agent Skill

Name: Verify
Author: NVZver

File contents

Trace. On load, print first: =============== [lsa/skills/verify/SKILL.md] [lsa] ===============

LSA Verify (before — grounding)

See CORE.md §6 (the two checks). This is the before check; reconcile is the after check.

Role

Grounding checker.

Goal

Confirm the spec is grounded in real code and buildable, before any handoff.

Inputs

Input	Source
The spec (`requirements.md`, `<flow>.feature`)	`specify`
The codebase	`self`
Quality-gate checks (optional)	`.lsa.yaml` `gate:` — see `knowledge/quality-gate-contract.md`

Steps

Identify the modules / functions / types the spec names (your judgment — never scripted), then pass them as arguments to bash scripts/resolve-refs.sh <symbol>… and cite its per-symbol resolution (exists @ file:line | new | MISSING | OUT-OF-RANGE) as the reference map — instead of multiple manual searches. Resolving each symbol is deterministic lookup (.lsa/VISION.md §2 principle 10); the GROUNDED / NOT-GROUNDED judgment stays yours. (→ reference map)
For each user flow: confirm it is buildable on what exists; infeasible → flag. (→ feasibility)
Confirm every claim is cited and every [ASSUMPTION] is visible. (→ grounding verdict)
Where .lsa.yaml defines a gate: block, run each configured check and cite its command + exit code as the grounding evidence — do not re-derive the checks by hand. Run the block in one pass where the repo provides an aggregate runner (this repo: bash scripts/gate.sh, which reads the gate: block and prints each check's command + exit), and cite its consolidated output; absent a runner, run each configured command. A non-zero exit is a real defect (a broken citation, dangling link, or violated invariant) and yields NOT-GROUNDED. (→ gate results)

Output

GROUNDED or NOT-GROUNDED with grounding.md (per reference: exists @ file:line | new | [ASSUMPTION]), the cited gate: command + exit for each configured check, and any blockers.

Constraints

Never delegate an ungrounded spec. A NOT-GROUNDED verdict blocks delegate.
A non-zero gate: check BLOCKS the GROUNDED verdict. The gate is the Rule-7 artifact (per knowledge/quality-gate-contract.md); "grounded" is claimed only with each command + exit 0 cited, never asserted.

/lsa:verify — manual invocation.

---name: verifydescription: Ground the spec against the codebase BEFORE delegating. Output — GROUNDED / NOT-GROUNDED + grounding.md.---> **Trace.** On load, print first: `=============== [lsa/skills/verify/SKILL.md] [lsa] ===============`# LSA Verify (before — grounding)See [CORE.md](../../CORE.md) §6 (the two checks). This is the *before* check; `reconcile` is the *after* check.## RoleGrounding checker.## GoalConfirm the spec is grounded in real code and buildable, before any handoff.## Inputs| Input | Source ||-------|--------|| The spec (`requirements.md`, `<flow>.feature`) | `specify` || The codebase | `self` || Quality-gate checks (optional) | `.lsa.yaml` `gate:` — see [`knowledge/quality-gate-contract.md`](../../knowledge/quality-gate-contract.md) |## Steps1. Identify the modules / functions / types the spec names (your judgment — never scripted), then pass them as arguments to `bash scripts/resolve-refs.sh <symbol>…` and cite its per-symbol resolution (`exists @ file:line` | `new` | `MISSING` | `OUT-OF-RANGE`) as the reference map — instead of multiple manual searches. Resolving each symbol is deterministic lookup (`.lsa/VISION.md` §2 principle 10); the GROUNDED / NOT-GROUNDED judgment stays yours. (→ reference map)2. For each user flow: confirm it is buildable on what exists; infeasible → flag. (→ feasibility)3. Confirm every claim is cited and every `[ASSUMPTION]` is visible. (→ grounding verdict)4. Where `.lsa.yaml` defines a `gate:` block, run each configured check and cite its command + exit code as the grounding evidence — do **not** re-derive the checks by hand. Run the block in **one pass** where the repo provides an aggregate runner (this repo: `bash scripts/gate.sh`, which reads the `gate:` block and prints each check's command + exit), and cite its consolidated output; absent a runner, run each configured command. A non-zero exit is a real defect (a broken citation, dangling link, or violated invariant) and yields `NOT-GROUNDED`. (→ gate results)## Output**GROUNDED** or **NOT-GROUNDED** with `grounding.md` (per reference: `exists @ file:line` | `new` | `[ASSUMPTION]`), the cited `gate:` command + exit for each configured check, and any blockers.## Constraints- **Never delegate an ungrounded spec.** A `NOT-GROUNDED` verdict blocks `delegate`.- **A non-zero `gate:` check BLOCKS the GROUNDED verdict.** The gate is the Rule-7 artifact (per [`knowledge/quality-gate-contract.md`](../../knowledge/quality-gate-contract.md)); "grounded" is claimed only with each command + exit `0` cited, never asserted.---`/lsa:verify` — manual invocation.

Frequently asked questions

Verify