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.
npx agentmods add commands/primeline-ai/universal-planning-framework/plan-refinegit clone --depth 1 https://github.com/primeline-ai/universal-planning-frameworkWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/primeline-ai/universal-planning-framework/plan-refine)<a href="https://agentmods.dev/commands/primeline-ai/universal-planning-framework/plan-refine"><img src="https://agentmods.dev/badge/commands/primeline-ai/universal-planning-framework/plan-refine.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00016 | $0.01667 |
| Opus 5 | $0.00008 | $0.00834 |
| Sonnet 5 | $0.00003 | $0.00333 |
| Haiku 4.5 | $0.00002 | $0.00167 |
Grade A, and why
plan-refine scanned grade A with 0 findings 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 4d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 190 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Autonomous Plan Hardening (Stage 1.5)
You are hardening this plan: $ARGUMENTS
Read the plan file and stress-test it from 6 adversarial perspectives. Fix what you can, flag what needs user input. Do NOT ask the user any questions - work autonomously.
Pre-Flight Checks
- Read the plan file using the Read tool
- Check idempotency: If a "Hardening Log" section already exists, inform the user and stop (plan already hardened)
- Count phases: If < 3 phases, inform the user that hardening is designed for complex plans and offer to proceed anyway
- Detect domain: Determine which of the 8 domains apply:
- Software Development, Multi-Agent / AI System, Business / Strategy, Content / Marketing
- Infrastructure / DevOps, Data & Analytics, Research / Exploration, Multi-Domain
Mode Selection
Simple mode (default): Single-agent, sequential. You role-switch through all 6 perspectives yourself.
Team mode (--team flag): Only when CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 is set:
- Spawn 6 agents (one per perspective), each reads the ORIGINAL plan in parallel
- Collect findings, deduplicate, resolve conflicts
- Conservative fix wins. Pedantic Lawyer takes priority on gates.
- Apply fixes, run consolidation check, append Hardening Log
If --team is in the arguments but the env var is not set, fall back to simple mode with a note.
6 Perspectives (Simple Mode - Sequential)
For each perspective, read the plan through that lens. Record findings and fixes.
Perspective 1: Outside Observer
"I'm reading this plan for the first time with no context."
- Can I summarize the goal in 15 words or fewer?
- Are success metrics unambiguous? Would two people measure them the same way?
- Is there an End State paragraph? Can I picture what success looks like?
- Are all abbreviations and jargon defined?
- Does Context & Why explain WHY, not just WHAT?
Perspective 2: Pessimistic Risk Assessor
"Everything that can go wrong, will go wrong."
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.
- 4d ago First seen · 190 lines · 0 tokens per session scan A cbb092cb6508
plan-refine is a command published in the GitHub repository primeline-ai/universal-planning-framework (45 stars, last pushed 10d ago), licensed MIT. It adds 16 tokens to every session and 1,667 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
bug-council
Bug Council — spawn 5 diagnostic specialists in parallel to analyze a hard bug from multiple angles, then synthesize a root cause and fix recommendation.
plan
Plan from the current Fathom session, grounded in a structured intent.
discovery-ping
Discovery-plugin self-test. Exercises the plugin-shipped slash command path so we can capture commandsource/expansiontype values via the hook dumps.
validate
Run the FPF Reference MCP validation contract (index status + tool list + compact route query) and report layered evidence.
pr-check-and-fix
Triage failing PR checks and apply fixes using limps workflows and scripts only.
speckit.tasks
Generate an actionable, dependency-ordered tasks.md for the feature based on available design artifacts.