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 skills/theagenticguy/erpaval/ultraplannpx skills add theagenticguy/erpaval --skill ultraplangit clone --depth 1 https://github.com/theagenticguy/erpavalWrote 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/skills/theagenticguy/erpaval/ultraplan)<a href="https://agentmods.dev/skills/theagenticguy/erpaval/ultraplan"><img src="https://agentmods.dev/badge/skills/theagenticguy/erpaval/ultraplan.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 | $0.00101 | $0.00776 |
| Opus 5 | $0.00051 | $0.00388 |
| Sonnet 5 | $0.00020 | $0.00155 |
| Haiku 4.5 | $0.00010 | $0.00078 |
Grade A, and why
ultraplan 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 5d 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 — 53 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Contents
| Reference | When to load |
|---|---|
references/orchestrator.md |
Running /ultraplan <problem> — the 5-phase flow |
references/divergence-vectors.md |
The three explorer biases and their motivation |
references/write-protocol.md |
Copy verbatim into every explorer prompt |
templates/explorer-skeleton.md |
Per-explorer plan skeleton |
templates/plan-skeleton.md |
Final composed plan skeleton |
Ultraplan
Generator-critic planning, file-coordinated. Three Opus 4.7 explorers plan the same problem in parallel, each biased toward a different axis — architectural, speed-first, simple-first. A critic reads all three and composes a final plan that cherry-picks the best decision from each, with attribution. The user gets one clean plan.md to execute against.
Beats single-agent planning on two axes:
- No anchoring. A single agent commits to the first approach it reaches and defends it through the rest of the plan. Three independent explorers reach genuinely different conclusions, and the critic picks per-decision.
- Calibrated confidence. When all three converge, the approach is sound. When they diverge, the critic names the tradeoff explicitly.
How it works
Read references/orchestrator.md and follow its phases. Short version:
- Intake — parse the problem. One targeted
AskUserQuestiononly if scope is genuinely ambiguous; otherwise proceed. - Skeletons — create
planning/{slug}/with three explorer files and aplan.mdshell. - Launch — three
general-purposeOpus 4.7 Agents in parallel,run_in_background: true, each with a distinct divergence vector fromreferences/divergence-vectors.md. Orchestrator does not read the codebase; explorers do. - Monitor — escalating check-ins (30s → 2m → 5m).
wc -lon explorer files; stuck detection via line-count delta; launch a fresh Agent with "resume from section N" context on stall. - Critic synthesis — one foreground
general-purposeOpus 4.7 Agent reads all three explorer files and writesplan.md— chosen approach, per-decision attribution, explicit tradeoffs, step-by-step order, verification criteria.
What ships with it
5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 5d ago First seen · 53 lines · 101 tokens per session scan A e619817938f8
ultraplan is a skill published in the GitHub repository theagenticguy/erpaval (28 stars, last pushed 3mo ago), licensed MIT. It adds 101 tokens to every session and 776 once invoked, about $0.0005 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 skills, from other repositories
discover
Use when a contextualizer's reference coverage needs to grow against its registered sources — a fresh contextualizer's first pass, quarterly upkeep, or whenever the catalog is lagging what users are asking — to propose new reference files.
new-reference
Use when a single topic is already identified and a full DISCOVER pass would be overkill, to register one new reference in an existing contextualizer.
refresh
Use when an existing contextualizer's references may have drifted from current upstream state — typically weekly, or whenever a few days of upstream changes have accumulated — to bring them back into agreement.
review
Use when DISCOVER or REFRESH has surfaced a proposal staged at -context.proposed/, to inspect it and record sign-off before applying it.
status
Use when picking up a contextualizer after a gap, or checking reference freshness and pending review work at a glance — read-only, safe to run anytime.
using-skill-engine
When the user mentions skill-engine or "the engine" without naming a specific workflow, or wants first-run setup. Inspects .claude/skills/-context/ install state (and any pending -context.proposed/ proposals) across all three install levels, then dispatches to engine-bootstrap when no contextualizer exists, to…