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/djrhails/dotfiles/reasoning-transparencynpx skills add DJRHails/dotfiles --skill reasoning-transparencygit clone --depth 1 https://github.com/DJRHails/dotfilesWhat 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 | $0.00158 | $0.02975 |
| Opus 5 | $0.00079 | $0.01488 |
| Sonnet 5 | $0.00032 | $0.00595 |
| Haiku 4.5 | $0.00016 | $0.00298 |
Grade A, and why
reasoning-transparency 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 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.
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 — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Reasoning Transparency
How to write an analysis so the reader can answer the only question that matters when they read it: "How should I update my view in response to this?" Distilled from Reasoning Transparency (Luke Muehlhauser, Open Philanthropy / now Coefficient Giving, Dec 2017).
A reasoning-transparent document lets the reader cheaply judge: Is the evidence presented fairly or selectively? How much does the author know? What was the research process — what shortcuts were taken? How confident is the author in each claim, and on what basis? What's the key takeaway, and what would change the author's mind? Most scientific norms (methods sections, open data, COI statements) target this; the two things they routinely miss are per-claim confidence and per-claim support — which is where most of the cheap wins are.
The three recommendations (do these first)
- Open with a linked summary of key takeaways. Lead with the conclusions, each linking to the section that argues for it. If the summary can't link inline, follow it immediately with a linked table of contents. The reader should grasp the takeaways and drill into any one of them without reading linearly.
- Flag which considerations matter most. Say early — in the intro, or at the top of each section — which arguments or pieces of evidence are load-bearing for the conclusion. Don't make the reader reverse-engineer what's doing the work. (Bad: a report whose conclusion rests almost entirely on RCT evidence but never says so.)
- State confidence + support for every major claim. For each claim that's critical to your conclusion, indicate (a) how confident you are and (b) what kind of support you have. This is the highest-leverage habit and the rest of this skill operationalises it.
Calibrate effort to stakes — don't gold-plate
GiveWell's AMF review is the extreme model: 125 endnotes, support for nearly every claim, a sourced table with archived copies, an open-questions list, linked sub-reports. That level is more costly than it's worth for almost everything. The goal is to buy most of the transparency at a fraction of the cost — mostly via cheap inline phrasings (below) rather than exhaustive citation. Spend the effort on the load-bearing claims; signpost the rest quickly.
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.
- 2d ago First seen · 120 lines · 158 tokens per session scan A 2aa902f9564d
reasoning-transparency is a skill published in the GitHub repository DJRHails/dotfiles (2 stars, last pushed 8d ago), licensed MIT. It adds 158 tokens to every session and 2,975 once invoked, about $0.0008 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
autopilot-batch
Fan out a batch of autopilot-queued issues to parallel background worktree subagents — each runs /autopilot at the build model from its 'model:' label — with a gating review at Opus 5 or above and never below the build (Opus reviews Sonnet and Opus builds, Fable reviews Fable builds).
todoist-cli
Manage Todoist tasks, projects, labels, filters, sections, comments, reminders, and workspaces via the td CLI. Use when the user wants to view, create, update, complete, or organize Todoist items, or mentions tasks, inbox, today, upcoming, projects, labels, or filters.
autopilot
Carry a well-scoped GitHub issue through the full dev loop autonomously, stopping at a per-run tier boundary (PR-ready, or merge+deploy for small reversible changes).
qa-handoff
Generate a hands-on QA testing guide as a self-contained HTML page — for Rails apps or static (Hugo) sites. --publish uploads the HTML to the project's configured QA host.
walkthrough
Generate a hands-on browser walkthrough of a PR's user-facing changes to exercise before review; --publish posts the final version to the PR for QA.
dustoff
Assess a project that's been dormant for months and produce a prioritized re-entry plan — lifecycle stage, what's stale or broken, and how far the project's tooling has drifted from current conventions — delegating every fix to the skill that owns it, and optionally capturing the plan as a tracking issue that survives…