How do I install the Validate Plan skill?

Run npx skillmds@latest add garyld1962/validate-plan-6 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 Validate Plan 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 Validate Plan?

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 Validate Plan free to use?

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

Who published Validate Plan?

garyld1962 (@garyld1962) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Validate Plan by garyld1962 — Agent Skill

Name: Validate Plan
Author: garyld1962

Validate Plan

Validate a task graph plan's YAML, dependencies, write ownership and acceptance before execution, then review semantic readiness. Use for plan checks; do not execute or rewrite the plan unless requested.

garyld1962 b241a50 9 files · 38.2 KB Updated Sep 17, 2026

File contents

Validate plan

Read the plan contract. The bundled scripts require Python 3 and PyYAML; missing dependencies are a reported blocker.

Resolve the explicit plan path. If omitted, inspect execution-plan candidates in docs/plans and choose only an unambiguous match; do not pick a PRD by modification time.
Run python3 <this-skill>/scripts/validate_plan.py <plan.md>. Use --json for parsed task metadata, diagnostics and the exact-byte plan hash.
Review what the script cannot prove: observable/relevant acceptance, repo command validity, actual file ownership, requirement coverage, consistent closed decisions, referenced decision records and unresolved product choices.
Return PASS only when both structural and semantic checks pass. Distinguish structural success from readiness. Return all failures with task IDs and remedies.
When called from an authorized planning/execution workflow, repair scoped issues and revalidate; when asked only for review, return findings without changing files.

The execution, requirements and reporting references provide the shared downstream contract. Before a final execution success claim run python3 <this-skill>/scripts/validate_report.py <report.json> --plan <plan.md>.

Examples

Two independent tasks own the same lockfile → FAIL; assign an owner and dependency.
A syntactically valid plan says "the feature works" → semantic FAIL; require proof.

Closed decisions and open decisions

Validate the supplied decisions for conflicts and source authority. Do not reopen them merely because another design is possible.

Do not

Do not run acceptance commands during a read-only validation, equate script exit 0 with complete readiness, or schedule a malformed dependency graph with --force.

Codex integration

Use $validate-plan and native Codex tools. Read AGENTS.md; preserve current host permissions. Private agent briefs live under execute-plan/references/agent-prompts. Legacy references are archival and are not part of the active contract.

---name: validate-plan-6description: Validate a task graph plan's YAML, dependencies, write ownership and acceptance before execution, then review semantic readiness. Use for plan checks; do not execute or rewrite the plan unless requested.---# Validate planRead [the plan contract](references/plan-format.md). The bundled scripts requirePython 3 and PyYAML; missing dependencies are a reported blocker.1. Resolve the explicit plan path. If omitted, inspect execution-plan candidates in docs/plans and choose only an unambiguous match; do not pick a PRD by modification time.2. Run `python3 <this-skill>/scripts/validate_plan.py <plan.md>`. Use --json for parsed task metadata, diagnostics and the exact-byte plan hash.3. Review what the script cannot prove: observable/relevant acceptance, repo command validity, actual file ownership, requirement coverage, consistent closed decisions, referenced decision records and unresolved product choices.4. Return PASS only when both structural and semantic checks pass. Distinguish structural success from readiness. Return all failures with task IDs and remedies.5. When called from an authorized planning/execution workflow, repair scoped issues and revalidate; when asked only for review, return findings without changing files.The [execution](references/execution.md), [requirements](references/prd-planning.md)and [reporting](references/reporting.md) references provide the shared downstreamcontract. Before a final execution success claim run`python3 <this-skill>/scripts/validate_report.py <report.json> --plan <plan.md>`.## Examples- Two independent tasks own the same lockfile → FAIL; assign an owner and dependency.- A syntactically valid plan says "the feature works" → semantic FAIL; require proof.## Closed decisions and open decisionsValidate the supplied decisions for conflicts and source authority. Do not reopen themmerely because another design is possible.## Do notDo not run acceptance commands during a read-only validation, equate script exit 0with complete readiness, or schedule a malformed dependency graph with --force.## Codex integrationUse $validate-plan and native Codex tools. Read AGENTS.md; preserve current host permissions.Private agent briefs live under execute-plan/references/agent-prompts. Legacy referencesare archival and are not part of the active contract.

Frequently asked questions