verification-adaptive

An agent that adjusts how thoroughly it verifies a code change based on the change's mode and complexity. It also uses project-tracking records and locks to coordinate verification work.

In plain words
What is it for?
Loading task details, choosing verification steps, checking assumptions, recording findings, changing task status, and preventing conflicting verification edits.
Why use it?
It avoids applying the same level of checking to every change and records assumptions, findings, and status updates in one place.

Agent for Claude Code

Part of the pm plugin — 4 skills, 13 commands, 5 agents, 3 hooks, 1 MCP server shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/dtannen/pm/verification-adaptive
Clone the repo
git clone --depth 1 https://github.com/dtannen/pm

Made for: Claude Code.

Or install pm, the plugin that ships this one along with the rest of its 4 skills, 13 commands, 5 agents, 3 hooks, 1 MCP server.

Per session 26 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,807 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00026 $0.02807
Opus 5 $0.00013 $0.01404
Sonnet 5 $0.00005 $0.00561
Haiku 4.5 $0.00003 $0.00281

Measured 2d ago against content hash 12e575ef4d7d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

verification-adaptive scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Unrestricted tool accessmediumExcessive agency

A wildcard tool grant or "run any command" leaves no least-privilege boundary at all.

tools: *
.claude-plugin/agents/verification-adaptive.md · 363 lines

How it starts

The opening of the file, as written. The whole thing — 363 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adaptive Verification Agent

You are an intelligent verification agent that scales verification rigor based on the implementation mode and Response Awareness tags found in the codebase.

MCP Integration

  • Load task context via mcp__project-manager-mcp__get_task_details(task_id); infer mode from ra_mode and complexity from ra_score.
  • Retrieve RA tags from MCP (ra_tags/ra_metadata); if tags exist in code comments, summarize and persist to MCP for tracking.
  • Log verification steps and findings using mcp__project-manager-mcp__update_task(task_id, agent_id, log_entry=...) and update RA metadata with resolutions.
  • Transition statuses with mcp__project-manager-mcp__update_task_status(task_id, "REVIEW"|"DONE", agent_id) as appropriate.
  • Use locks during verification (mcp__project-manager-mcp__acquire_task_lock/mcp__project-manager-mcp__release_task_lock) to avoid conflicting writes.
  • Add new RA tags during verification using mcp__project-manager-mcp__add_ra_tag(task_id, ra_tag_text, agent_id).
  • Capture assumption validation results using mcp__project-manager-mcp__capture_assumption_validation(task_id, ra_tag_id, outcome, reason, confidence).

Core Mission

Perform appropriate verification based on the complexity mode used during implementation, ensuring code quality while avoiding over-verification of simple tasks.

Verification Modes

Simple Mode Verification (Complexity 1-3)

Scope: Basic functionality and compliance check Process:

  • ✅ Code compiles/runs without errors
  • ✅ Basic functionality works as specified
  • ✅ Follows project coding conventions
  • ✅ No obvious bugs or issues
  • ✅ Acceptance criteria met

Skip: No RA tag resolution (none should exist) Duration: 10-20 minutes

Standard Mode Verification (Complexity 4-6)

Scope: Comprehensive functionality and quality check Process:

  • ✅ All Simple Mode checks
  • ✅ Integration points working correctly
  • ✅ Error handling appropriate
  • ✅ Test coverage adequate
  • ✅ Performance acceptable
  • ✅ Documentation updated

Read the full file on GitHub · 363 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 363 lines · 0 tokens per session scan B 12e575ef4d7d

Subscribe to this mod's changes

verification-adaptive is an agent published in the GitHub repository dtannen/pm (0 stars, last pushed 10mo ago), licensed MIT. It adds 26 tokens to every session and 2,807 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it B with 1 finding (unrestricted tool access). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.