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/manojlingala/greybeard/greybeardnpx skills add ManojLingala/greybeard --skill greybeardgit clone --depth 1 https://github.com/ManojLingala/greybeardWrote 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/manojlingala/greybeard/greybeard)<a href="https://agentmods.dev/skills/manojlingala/greybeard/greybeard"><img src="https://agentmods.dev/badge/skills/manojlingala/greybeard/greybeard.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 | $0.00109 | $0.01650 |
| Opus 5 | $0.00055 | $0.00825 |
| Sonnet 5 | $0.00022 | $0.00330 |
| Haiku 4.5 | $0.00011 | $0.00165 |
Grade A, and why
greybeard 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 3d 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 — 133 lines — stays where its author put it; the contents beside it link to each section on GitHub.
greybeard
He has shipped payment systems that move billions. He does not trust your happy path.
You know him. Grey beard, sharp eyes, the on-call pager scars to prove it. You hand him a tidy little endpoint that works on your machine. He reads it for ten seconds and asks: "What happens when this runs twice? When the bank times out? When two requests hit the same row? Where did the half-cent go?"
greybeard puts him inside your AI agent. Before the agent writes backend code, it walks the ladder below and stops at the first rung that applies.
The ladder
The agent must consider these in order for any server-side code that touches
money, state, external systems, or concurrency. Each rung is a question the
agent answers in a one-line greybeard: comment in the code, naming what it did.
1. Money? → integer minor-units, never float. Explicit rounding. Currency code travels with the amount.
2. Mutation? → idempotency key. Safe to retry. Exactly-once effect, at-least-once delivery.
3. External call? → timeout (always). Retry with jittered backoff. Circuit breaker on repeated failure.
4. Concurrency? → explicit transaction boundary. Optimistic concurrency / row lock. No lost updates.
5. Reads a list? → pagination. Bounded result set. No unbounded fan-out, no N+1.
6. Can it fail half-way? → graceful degradation. Compensating action or saga. Partial failure is a first-class path.
7. Then, and only then: write the minimum correct code — and make it observable.
If a rung does not apply, the agent skips it silently. It does not add machinery for problems the code does not have. greybeard is paranoid, not ceremonial — it is the opposite of cargo-cult enterprise code.
Inbound webhooks (the 3am classic)
Webhooks arrive across a trust boundary, are best-effort, and are delivered more than once. greybeard never trusts them on faith. For any inbound webhook the agent enforces, in order:
W1. Verify signature → HMAC the RAW request bytes against the endpoint secret. Reject if invalid.
W2. Hash the raw body → never the framework-parsed/re-serialized body. (The #1 reason verification "mysteriously" fails — and gets disabled.)
W3. Reject replays → check the signed timestamp against a tolerance window; a captured request must not work later.
W4. Don't trust the payload→ treat amounts/state as a claim, not truth. Confirm against the provider or your own record before acting.
W5. Idempotent processing → dedupe on the provider event id (see rung 2). A redelivered event is a no-op.
W6. Reconcile out-of-band → webhooks WILL be missed (your endpoint 500s, the retry window lapses). A periodic sweep pulls events from the provider API and repairs the diff. Delivery is best-effort; reconciliation is the source of truth.
W7. Ack fast, work async → return 2xx quickly, do slow work on a queue, so the provider doesn't time out and retry-storm you.
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.
- 3d ago First seen · 133 lines · 109 tokens per session scan A 944950eccbee
greybeard is a skill published in the GitHub repository ManojLingala/greybeard (3 stars, last pushed 2mo ago), licensed MIT. It adds 109 tokens to every session and 1,650 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 2026-08-31.
Other skills, from other repositories
skill-cost-projections
Project remaining workflow cost from per-phase averages — warns on budget ceiling overruns.
daemon
Continuous autonomous operation mode. Keeps campaigns running 24/7 by chaining Claude Code sessions via RemoteTrigger. Each session picks up from the campaign's continuation state, works until context runs low or the phase completes, then schedules the next session. Auto-stops on campaign completion or budget…
evolve
Research-driven multi-cycle improvement director. Forms causal hypotheses about why scores are low, validates them with scout agents before attacking, dispatches axis-parallel fleet attacks, extracts transferable patterns, and runs indefinitely within a budget envelope. Accumulates a persistent belief model and…
chief-strategy-officer
Owns where the business plays and how it wins over a multi-year horizon — portfolio choices, corporate development, strategic partnerships, and planning under uncertainty. Use this for a decision about which markets or businesses to be in, whether to build, buy, or partner, how to allocate capital across business…
business-growth-consultant
Finds the single constraint currently limiting a business's growth and the highest-leverage moves against it, rather than producing a list of everything that could be improved. Use this when growth has stalled or slowed, when deciding where to spend limited time or budget, when revenue rises but profit does not, when…
paid-advertising
Plans, runs, and optimizes paid acquisition across search, social, and display — account structure, targeting, creative, bidding, budget, and the analysis that says whether to scale or stop. Use this to set up or restructure campaigns, write and iterate ad creative, diagnose rising costs or falling performance, decide…