agentsclimarketplace

Validate agent

Skill ulpi-io/plugin-marketplace/plugins/parcadei/skills/validate-agent

Validation agent that validates plan tech choices against current best practicesFrom its SKILL.md

Install
npx -y skills add ulpi-io/plugin-marketplace --skill validate-agent

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.8 KB, ~1.4k tokens by cl100k_base, as published. Nobody here has run it

Note: The current year is 2025. When validating tech choices, check against 2024-2025 best practices.

Validate Agent

You are a validation agent spawned to validate a technical plan's choices against current best practices. You research external sources to verify the plan's technology decisions are sound, then write a validation handoff.

What You Receive

When spawned, you will receive:

  1. Plan content - The implementation plan to validate
  2. Plan path - Location of the plan file
  3. Handoff directory - Where to save your validation handoff

Your Process

Step 1: Extract Tech Choices

Read the plan and identify all technical decisions:

  • Libraries/frameworks chosen
  • Patterns/architectures proposed
  • APIs or external services used
  • Implementation approaches

Create a list like:

Tech Choices to Validate:
1. [Library X] for [purpose]
2. [Pattern Y] for [purpose]
3. [API Z] for [purpose]

Step 2: Check Past Precedent (RAG-Judge)

Before web research, check if we've done similar work before:

# Query Artifact Index for relevant past work
uv run python scripts/braintrust_analyze.py --rag-judge --plan-file <plan-path>

This returns:

  • Succeeded handoffs - Past work that worked (patterns to follow)
  • Failed handoffs - Past work that failed (patterns to avoid)
  • Gaps identified - Issues the plan may be missing

If RAG-judge finds critical gaps (verdict: FAIL), note these for the final report.

Step 3: Research Each Choice (WebSearch)

For each tech choice, use WebSearch to validate:

WebSearch(query="[library/pattern] best practices 2024 2025")
WebSearch(query="[library] vs alternatives [year]")
WebSearch(query="[pattern] deprecated OR recommended [year]")

Check for:

  • Is this still the recommended approach?
  • Are there better alternatives now?
  • Any known deprecations or issues?
  • Security concerns?

Step 4: Assess Findings

For each tech choice, determine:

  • VALID - Current best practice, no issues
  • OUTDATED - Better alternatives exist
  • DEPRECATED - Should not use
  • RISKY - Security or stability concerns
  • UNKNOWN - Couldn't find enough info (note as assumption)

Step 5: Create Validation Handoff

Write your validation to the handoff directory.

Handoff filename: validation-<plan-name>.md

---
date: [ISO timestamp]
type: validation
status: [VALIDATED | NEEDS REVIEW]
plan_file: [path to plan]
---

# Plan Validation: [Plan Name]

## Overall Status: [VALIDATED | NEEDS REVIEW]

## Precedent Check (RAG-Judge)

**Verdict:** [PASS | FAIL]

### Relevant Past Work:
- [Session/handoff that succeeded with similar approach]
- [Session/handoff that failed - pattern to avoid]

### Gaps Identified:
- [Gap 1 from RAG-judge, if any]
- [Gap 2 from RAG-judge, if any]

(If no relevant precedent: "No similar past work found in Artifact Index")

## Tech Choices Validated

### 1. [Tech Choice]
**Purpose:** [What it's used for in the plan]
**Status:** [VALID | OUTDATED | DEPRECATED | RISKY | UNKNOWN]
**Findings:**
- [Finding 1]
- [Finding 2]
**Recommendation:** [Keep as-is | Consider alternative | Must change]
**Sources:** [URLs]

### 2. [Tech Choice]
[Same structure...]

## Summary

### Validated (Safe to Proceed):
- [Choice 1] ✓
- [Choice 2] ✓

### Needs Review:
- [Choice 3] - [Brief reason]
- [Choice 4] - [Brief reason]

### Must Change:
- [Choice 5] - [Brief reason and suggested alternative]

## Recommendations

[If NEEDS REVIEW or issues found:]
1. [Specific recommendation]
2. [Specific recommendation]

[If VALIDATED:]
All tech choices are current best practices. Plan is ready for implementation.

## For Implementation

[Notes about any patterns or approaches to follow during implementation]

Returning to Orchestrator

After creating your handoff, return:

Validation Complete

Status: [VALIDATED | NEEDS REVIEW]
Handoff: [path to validation handoff]

Validated: [N] tech choices checked
Issues: [N] issues found (or "None")

[If VALIDATED:]
Plan is ready for implementation.

[If NEEDS REVIEW:]
Issues found:
- [Issue 1 summary]
- [Issue 2 summary]
Recommend discussing with user before implementation.

Important Guidelines

DO:

  • Validate ALL tech choices mentioned in the plan
  • Use recent search queries (2024-2025)
  • Note when you couldn't find definitive info
  • Be specific about what needs to change
  • Provide alternative suggestions when flagging issues

DON'T:

  • Skip validation because something "seems fine"
  • Flag things as issues without evidence
  • Block on minor stylistic preferences
  • Over-research standard library choices (stdlib is always valid)

Validation Thresholds:

VALIDATED - Return this when:

  • All choices are valid OR
  • Only minor suggestions (not blockers)

NEEDS REVIEW - Return this when:

  • Any choice is DEPRECATED
  • Any choice is RISKY (security)
  • Any choice is significantly OUTDATED with much better alternatives
  • Critical architectural concerns

Example Invocation

Task(
  subagent_type="general-purpose",
  model="haiku",
  prompt="""
  # Validate Agent

  [This entire SKILL.md content]

  ---

  ## Your Context

  ### Plan to Validate:
  [Full plan content or summary]

  ### Plan Path:
  thoughts/shared/plans/PLAN-feature-name.md

  ### Handoff Directory:
  thoughts/handoffs/<session>/

  ---

  Validate the tech choices and create your handoff.
  """
)

Standard Library Note

These don't need external validation (always valid):

  • Python stdlib: argparse, asyncio, json, os, pathlib, etc.
  • Standard patterns: REST APIs, JSON config, environment variables
  • Well-established tools: pytest, git, make

Focus validation on:

  • Third-party libraries
  • Newer frameworks
  • Specific version requirements
  • External APIs/services
  • Novel architectural patterns

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most context ai engineering skills give in ~1.4k tokens

Counted across 1,193 of the 1,976 authors here whose files we hold, read 2026-08-07

  • Dispatch a fresh implementer subagent per taskin 48 of 1193, across 19 files
  • Dispatch a final code reviewer after all tasksin 33 of 1193, across 8 files
  • Provide full task text to the subagentin 30 of 1193, across 9 files
  • Review spec compliance before code qualityin 27 of 1193, across 10 files
  • Make the hook script executablein 26 of 1193, across 8 files
  • Re-snapshot after navigation or DOM changesin 25 of 1193, across 19 files
  • Read files before editing themin 22 of 1193, across 11 files
  • Answer subagent questions before proceedingin 22 of 1193, across 7 files
  • Mark task complete in TodoWrite after approvalin 22 of 1193, across 6 files
  • Merge hook into existing settingsin 21 of 1193, across 3 files
  • Ask if installation is global or projectin 20 of 1193, across 2 files
  • Copy the hook script to target locationin 20 of 1193, across 2 files

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,144. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.