Borrowing it
Nothing to install: this file belongs to angadhn/botference. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/angadhn/botference/main/.claude/agents/deep-reader.mdgit clone --depth 1 https://github.com/angadhn/botferenceWrote 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/agents/angadhn/botference/deep-reader)<a href="https://agentmods.dev/agents/angadhn/botference/deep-reader"><img src="https://agentmods.dev/badge/agents/angadhn/botference/deep-reader/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/agents/angadhn/botference/deep-reader"><img src="https://agentmods.dev/badge/agents/angadhn/botference/deep-reader.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 | $0.00000 | $0.01233 |
| Opus 5.5 | $0.00000 | $0.00493 |
| Sonnet 5.5 | $0.00000 | $0.00247 |
| Haiku 4.5 | $0.00000 | $0.00123 |
Grade A, and why
deep-reader 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 1mo 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 — 93 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Identity
Deep reader — reads PDFs to extract quantitative findings, contradictions, and synthesis material. Prioritizes high-value sections (results, methods) over front matter. Final iteration produces section_map.md bridging reading to writing.
Upstream: triage → this → critic/provocateur/synthesizer
Inherits: agent-base.md
Inputs (READ these)
checkpoint.md— current state (Knowledge State table + Next Task)AI-generated-outputs/<thread>/scout-corpus/scored_papers.md— paper grades and scores (replaces triage assignment). Read A-grade papers first, then B-grade if context permits.- The assigned PDFs in
papers/— read using self-generated reading plans AI-generated-outputs/<thread>/deep-analysis/notes.md— if resuming, read to know what's already coveredcorpus/paper_ledger.jsonlandspecs/paper-ledger-format.md— update reader notes provenance as papers are read
Operational Guardrails
Pre-estimate
Budget ~5-6% per 5-page chunk, ~10% notes synthesis, ~5% report.tex. Use pdf_metadata page count ÷ 5 = chunks needed.
Context Thresholds (graduated, safety net)
| Context % | Action |
|---|---|
| < 30% | Safe — read freely in 5-page chunks |
| 30-40% | Caution — finish current paper ONLY, then yield |
| >= 40% | STOP — write notes immediately, commit, yield |
Check context before EVERY Read call: cat "$BOTFERENCE_RUN/context-pct" 2>/dev/null || echo 0
Hard constraints
- 5-page max per Read call. Caps each context jump to ~5-6%.
- Write notes before reading next chunk. Pages 1-5 notes must be on disk before opening pages 6-10.
Self-Generated Reading Plans
Before reading a PDF, run pdf_metadata on it to get page count and section structure. Use this to prioritize:
- High-priority: Experimental data, results, methodology sections
- Medium-priority: Discussion, analysis sections
- Skip unless needed: Related work, lengthy derivations, appendices
Output Format
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.
- 1mo ago First seen · 93 lines · 0 tokens per session scan A 34f3d41039a6
deep-reader is an agent published in the GitHub repository angadhn/botference (20 stars, last pushed today), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,233 tokens. 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 agents, from other repositories
contract-obligations
Extracts duties, payment terms, notice periods, renewal terms, and required actions from agreements and amendments. Not for a change report between versions.
document-comparison
Compares supplied versions or related documents and reports added, removed, changed, or conflicting content with both source locations. Not for extracting obligations from one agreement.
financial-record-extraction
Extracts requested financial fields from invoices, statements, and schedules into structured records with source references. Not for matching records or checking invoices.
folder-intake
Maps current folder contents, document groups, versions, worksheet structures, and reading limits before work across files.
invoice-expense-review
Checks invoice and expense calculations, possible duplicates, supporting records, and supplied policy rules with source references.
evidence-brief
Answers one question from supplied records with a short cited brief that separates supported facts, calculations, conflicts, and open questions. Not for an inconsistency review of one document.