ingest-content
Skill2 starsChanged 5 months agoMITView on GitHub
Ingest a document (PDF, article, white paper, book chapter, URL) and produce a Pyramid Principle analysis at multiple granularities — TL;DR, abstract, structural pillars, deep dive per section, key concepts. Optionally save to the operator's external knowledge tracker.
- Installs packages
In this repository24 other files it gives its agents.
- AGENTS.mdAGENTS.md
- control-plane/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- control-plane/CLAUDE.mdCLAUDE.md
- agentic-os-reference-syncSkill
- branch-cleanupSkill
- capture-triageSkill
- darwin-housekeepingSkill
- darwin-path-modeSkill
- decision-log-entrySkill
- eodSkill
- expert-interview-guideSkill
- feedback-intakeSkill
- harness-onboardingSkill
- meeting-to-work-itemsSkill
- memory-consolidateSkill
- morning-briefSkill
- os-bootstrap-extendSkill
- os-bootstrapSkill
- pr-reviewSkill
- scheduleSkill
- spec-cross-checkSkill
- ux-critiqueSkill
- weekly-retroSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 59
Raw text: use directly
- Very specific: line 100
If save_to_external=true and a tracker is configured: create a structured note in the operator's external knowledge system (Notion / …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Skill: ingest-content
Preconditions
- The input is a local PDF, a URL, or raw text — supplied via
sourceinput. pdfplumberis available if the source is a PDF (pip install pdfplumberonce).WebFetchis available for URL sources.- An external knowledge tracker (Notion, Obsidian, plain markdown) is configured if
save_to_external=true— otherwise output stays in the response and optionally inlearning/.
When to use
Operator hands you a document and wants it understood at depth, not just summarized. The output is structured so the principal can read the TL;DR in 5 seconds and the deep dive in 15 minutes, depending on need.
NOT for:
- A quick Q&A about a single fact in a document — just answer.
- Generating new content from research — that's a thought-leadership pipeline, not ingestion.
- Translating documents — different skill.
Input contract
source: str # local path (PDF) or URL or raw text
source_type: pdf | url | text
title: str | None # extracted from document if None
author: str | None
domain: str | None # which knowledge area (learning / professional / etc)
save_to_external: bool # default true if external tracker configured
context_hint: str | None # additional context (e.g. "relevant to X engagement")Output contract
tl_dr: str # 1-2 sentences — complete answer in minimum words
abstract: str # single ~150-200 word paragraph
argument_map: str # tree structure of the argument (markdown)
deep_dive: list[Section] # one per chapter/section
key_concepts: list[str] # 5-10 extracted concepts
source_notes: str # provenance, editorial bias, limitations
external_url: str | None # link to created note if saved externallySequence
Step 1 — Content extraction
- PDF local:
pdfplumberviapython3 -c "..."extracting page by page - URL:
WebFetchfor the content - Raw text: use directly
- Record: title, author / org, page or section count, chapter structure
Step 2 — TL;DR
1-2 sentences answering "what is the central thesis and why does it matter?". Must be intelligible without further context.
Step 3 — Abstract
~150-200 word paragraph structured as:
- Problem / context the document addresses
- Central thesis
- 2-3 supporting pillars
- Practical implication or call to action
Step 4 — Argument map
Tree structure of the argument:
THESIS: <central claim>
├── PILLAR 1: <support 1>
│ ├── Evidence A
│ └── Evidence B
├── PILLAR 2: <support 2>
└── PILLAR 3: <support 3>Step 5 — Deep dive per section
For each chapter / section, a ~100-150 word block that captures:
- What the section argues
- How it supports the central thesis
- Key data, examples, or definitions
- Friction or weakness if present (do not over-flatter)
Step 6 — Key concepts
5-10 named concepts the document introduces or relies on. These are vocabulary the principal will use when referencing the document later. Each concept gets a one-sentence definition.
Step 7 — Source notes
- Provenance (publisher, date, edition)
- Editorial bias (advocacy / academic / commercial / personal)
- Limitations (sample size, scope, time)
- Other documents this engages with or contradicts
Step 8 — Persist
- If
save_to_external=trueand a tracker is configured: create a structured note in the operator's external knowledge system (Notion / Obsidian / plain markdown inlearning/canon/). - Otherwise: return all sections inline.
Anti-patterns
- Summarizing without structure — defeats the Pyramid Principle. A summary is not the same as TL;DR + abstract + argument map.
- Skipping the source notes — provenance is what lets the operator weight the document's claims later.
- Treating every section as equally important — deep-dive blocks for trivial sections waste attention. Triage by load-bearing-ness.
- Hiding weaknesses — the operator wants the document understood honestly, including where it's thin.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

