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/lancegui/causal-powers/result-verificationnpx skills add lancegui/causal-powers --skill result-verificationgit clone --depth 1 https://github.com/lancegui/causal-powersWrote 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/lancegui/causal-powers/result-verification)<a href="https://agentmods.dev/skills/lancegui/causal-powers/result-verification"><img src="https://agentmods.dev/badge/skills/lancegui/causal-powers/result-verification.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 | $0.00176 | $0.03489 |
| Opus 5 | $0.00088 | $0.01744 |
| Sonnet 5 | $0.00035 | $0.00698 |
| Haiku 4.5 | $0.00018 | $0.00349 |
Grade A, and why
result-verification 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 — 128 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Result Verification
Overview
This skill is user-invoked. When a result is about to ship unverified, offer verification in one line — never run it unasked (and never widen it: see below).
The last mile is where good analyses die. The number is computed, it looks right, the deadline is close — and "looks right" becomes "is right" without anything in between. This skill is the something in between: the checks that stand between a computed number and a claimed result.
This is the analytics counterpart of verification-before-completion. The rule is identical: evidence before assertions, always. You do not say "the result is X"; you say "the result is X, here is the reconciliation, here is the clean-room reproduction, here is what the approved robustness showed." Verification confirms; it never widens — running a spec, cut, or placebo nobody approved is not extra diligence, it's an unapproved analysis decision (analysis-checkpoints).
Core principle: A result is not done when it appears; it's done when it has been reconciled, reproduced from scratch, and survived being attacked.
The verification checklist
Run these before any result leaves your hands. Each maps to a real way final numbers turn out wrong:
-
Reconcile to source. Do the parts sum back to the known whole? Does the headline number tie out to a total you can compute a completely different way? A revenue figure should reconcile to the raw ledger; a user count to a distinct count of IDs. Reconciliation by an independent path is the single strongest check that the number is real. (Use float-aware comparison.)
-
Reproduce from a clean state. Restart the kernel / session / R process — no cached objects, no leftover variables — set the seed, and run the analysis end to end from raw inputs. A result that only exists because of a variable still in memory from three hours ago is not a result. If it doesn't reproduce, you don't have a finding, you have an artifact.
-
Confirm determinism. Same input + same seed → same output, twice. If the number wiggles between runs, there's uncontrolled randomness or ordering dependence, and the figure you're about to report is one sample from a distribution you didn't mean to draw from.
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 · 128 lines · 176 tokens per session scan A 5afe9fdaefd8
result-verification is a skill published in the GitHub repository lancegui/causal-powers (2 stars, last pushed 10d ago), licensed MIT. It adds 176 tokens to every session and 3,489 once invoked, about $0.0009 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
audit-reproducibility
Enforce the replication-protocol.md rule by cross-checking numeric claims in a manuscript against the actual R / Stata / Python outputs. Report PASS/FAIL per claim against tolerance thresholds. Use before submission and before releasing a replication package.
diagnose
Root-cause a failing or wrong empirical result with a disciplined reproduce → minimise → hypothesise → instrument → fix loop, instead of guessing-and-poking. Use when the user says "why is my regression wrong", "this number changed", "my script errors out", "the result won't reproduce", "debug this", "this estimate…
capture-environment
Snapshot the computational environment for a replication package — detects the analysis stack (R / Stata / Python) and emits the right lockfiles (renv.lock + sessionInfo.txt, requirements.txt / environment.yml / uv.lock, Stata version + ado package list), records seeds and RNG kind, optionally writes a pinning…
checkpoint
Save a structured state snapshot before stopping or handing off. Captures the active plan, recent decisions, file pointers (with line numbers), open questions, and the next 1–3 actions into a checkpoint file under qualityreports/checkpoints/. Optionally proposes [LEARN] entries to add to MEMORY.md. Use when user says…
coauthor-brief
Generate a co-author / collaborator handoff brief for a multi-author, multi-machine project — summarizing what changed since the last brief (git delta), the current state of each artifact (manuscript, analysis, slides), open questions, how to reproduce locally, and any restricted-data access steps. Use when user says…
data-management-plan
Draft a funder-compliant Data Management Plan (NSF DMP, NIH DMS Policy 2023, ERC, Horizon Europe) by composing the confidential-data and environment-capture primitives. Sections cover data description, formats/metadata, storage/backup, access/sharing, preservation/archiving, and roles. Use when user says "data…