Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/roomi-fields/paper-trailnpx agentmods add commands/roomi-fields/paper-trail/paper-trail-decideWrote 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/roomi-fields/paper-trail/paper-trail-decide)<a href="https://agentmods.dev/commands/roomi-fields/paper-trail/paper-trail-decide"><img src="https://agentmods.dev/badge/commands/roomi-fields/paper-trail/paper-trail-decide/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/roomi-fields/paper-trail/paper-trail-decide"><img src="https://agentmods.dev/badge/commands/roomi-fields/paper-trail/paper-trail-decide.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00042 | $0.00740 |
| Opus 5 | $0.00021 | $0.00370 |
| Sonnet 5 | $0.00008 | $0.00148 |
| Haiku 4.5 | $0.00004 | $0.00074 |
Grade A, and why
paper-trail-decide 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 9d 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 — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/paper-trail:decide <slug> — Inspect & decide on one ref
Pour une ref précise : montre tout ce qui est nécessaire pour décider, et applique la décision.
Usage
/paper-trail:decide shannon_1948_mathematical_theory
/paper-trail:decide # liste les refs problématiques d'abord
Ce que fait Claude
-
Récupérer toute l'info de la ref via :
python3 tools/review_problems.py --ref-slug <slug> --output -(ou affiche directement le contenu du fichier registre
_registry/refs/<slug>.md- scan des citations dans le vault si besoin)
-
Présenter en français concis :
- Identité : auteur, année, titre, UID
- État actuel + raison du blocage
- PDF : présent sur disque ? sha ? page 1 validée ?
- Sources tentées : liste des sources cascade avec verdict (no_source / failed / page1_failed / success)
- Citations dans le vault : dans quels SOTAs/articles cette ref est-elle mentionnée ? avec extrait de contexte
- Historique d'états (state_history) si non trivial
-
Demander la décision via AskUserQuestion. Les options proposées dépendent de l'état :
-
Si
blocked_human:cascade_exhausted:- retract : la ref n'a pas d'importance, on l'écarte
- unblock : retenter la cascade (utile si tu as ajouté un proxy, ou changé la config réseau)
- investigate : tu vas corriger l'auteur/titre/uid dans le frontmatter
- keep : laisser comme ça (no-op)
-
Si
uid_resolved(cascade épuisée mais pas encore blocked) :- retract / blocked / investigate / keep
-
Si
awaiting_rtfm_ocr:- retract / wait (laisser RTFM finir l'OCR) / investigate
-
-
Appliquer la décision via :
python3 -m pipeline arbitrate <slug> --decision <X> --reason "<reason>" -
Confirmer : afficher le nouvel état de la ref.
Style
- Concis, termes du domaine (recherche / musicologie), pas de jargon FSM.
- "État actuel" plutôt que "state", "sources tentées" plutôt que "acquisition_attempts".
- Pour les citations : extrait de ±2 lignes autour du
[[slug]]pour voir le contexte (de quoi parle la SOTA quand elle cite cette ref). - Toujours demander avant d'appliquer une décision destructrice (retract).
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.
- 9d ago First seen · 73 lines · 42 tokens per session scan A 04d8342b47d9
paper-trail-decide is a command published in the GitHub repository roomi-fields/paper-trail (5 stars, last pushed 5d ago), licensed MIT. It adds 42 tokens to every session and 740 once invoked, about $0.0002 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 commands, from other repositories
gh-release
Create a new release for the marketplace or a specific plugin.
commit
Classify changes and create atomic commits with conventional messages. Flags out-of-context modifications and red-flag patterns before committing.
resolve
Resolve merge conflicts on the current branch or for a pull request. Detects conflict types, analyzes both sides, applies per-file resolution strategies, and verifies the result compiles and passes tests.
ledger-research
Launch parallel evidence collection across all active research pillars.
ledger-spec
Generate constrained PRD and architecture documents where every section cites decisions.
quick-research
Quick research with source verification (lighter than /deep-research).