ingest

ingest is a skill for Claude Code, Codex from xcota/pos. It costs 19 tokens per session (1,040 once invoked), scanned A, original, MIT.

A command for extracting structured facts from raw files or sources into a knowledge graph, a connected collection of linked information.

In plain words
What is it for?
Use it to ingest Telegram data, transcripts, YouTube subtitles, notes, biographies, surveys, or audio, optionally focusing on a specific topic.
Why use it?
It turns unorganized material such as transcripts, notes, surveys, biographies, and audio into information that can be searched and connected.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/xcota/pos/ingest
Any agent
npx skills add xcota/pos --skill ingest
Clone the repo
git clone --depth 1 https://github.com/xcota/pos

Made for: Claude Code, Codex.

Wrote 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.

agentmods badge for ingest

README.md
[![agentmods](https://agentmods.dev/badge/skills/xcota/pos/ingest.svg)](https://agentmods.dev/skills/xcota/pos/ingest)
Your own site
<a href="https://agentmods.dev/skills/xcota/pos/ingest"><img src="https://agentmods.dev/badge/skills/xcota/pos/ingest.svg" alt="Measured on agentmods" height="20"></a>
Per session 19 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,040 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00019 $0.01040
Opus 5 $0.00010 $0.00520
Sonnet 5 $0.00004 $0.00208
Haiku 4.5 $0.00002 $0.00104

Measured 4d ago against content hash 1af28baa6571, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ingest 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 4d 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.

.claude/skills/ingest/SKILL.md · 59 lines

How it starts

The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/ingest — Structured Data Extraction

Extract structured knowledge from raw data sources into the knowledge graph.

Arguments

  • [source_type] — source kind: telegram | transcript | youtube | notes | bio | survey | audio
  • [path] — path to the file or directory holding the data
  • [topic] — optional topic/context to focus extraction

Pipeline

Phase 1: R2C (Raw → Characteristics)

  1. Detect source type (auto-detect, or from the argument).
    • YouTube URL: create inbox/youtube/{slug}_{video_id}/, fetch auto-subs with yt-dlp --skip-download --write-auto-subs --sub-langs "en.*,<primary>.*" --sub-format vtt (use the subject's primary language from context/identity.md if profiled, else en; run yt-dlp --list-subs <url> first if unsure). Needs yt-dlp installed. If no .vtt is produced (yt-dlp exits 0 even when it finds nothing), do NOT fabricate a transcript — mark the source unresolved and ask the user to paste one. Otherwise normalize VTT into transcript.md + transcript.txt, then process as transcript.
  2. Load the source-specific extraction prompt from projects/ingest/prompts/{source_type}.md.
    • If projects/ingest/prompts/youtube.md is missing, use transcript.md and add focus on: source metadata, chapter outline, reusable concepts, architecture deltas, action items.
  3. Split input into chunks (~1500 tokens, with overlap).
  4. Parallel subagents: each chunk × each dimension —
    • bio — biographical facts, events, chronology
    • patterns — patterns of behavior, decisions, thinking
    • values — values, beliefs, priorities
    • connections — people, relationships, social graph
    • insights — insights, unique ideas, non-obvious observations
  5. Output format per chunk:
    date/period – dimension – fact/observation – "evidence quote" – confidence (0-1)
    

Phase 2: C2F (Characteristics → Final)

  1. Collect all Phase 1 results.
  2. Deduplicate: merge identical facts, raise confidence.
  3. Consolidate: group by topic/entity.
  4. Create/update files in knowledge/:
    • New entity → knowledge/{type}/{entity-name}.md with frontmatter + [[wikilinks]].
    • Existing entity → append/merge new facts.
    • Wikilink validation: every [[link]] must point at a file that exists.

Read the full file on GitHub · 59 lines

Changes

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.

  1. 4d ago First seen · 59 lines · 19 tokens per session scan A 1af28baa6571

Subscribe to this mod's changes

ingest is a skill published in the GitHub repository xcota/pos (43 stars, last pushed 1mo ago), licensed MIT. It adds 19 tokens to every session and 1,040 once invoked, about $0.0001 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 2026-08-30.

Related

Other skills, from other repositories

ori-memory

Persistent agent memory with learning retrieval. Knowledge graph on markdown files — capture insights, decisions, research, and learnings during work, then retrieve them weeks or months later. Use when knowledge is too valuable to lose but too much to inject into every prompt.

aayoawoyemi/Ori-Mnemos · 54 tokens

rtg-knowledge-graph-mcp

Operate the Vellis RTG Knowledge Graph MCP server for agent-driven graph work. Use when an agent has RTG MCP tools such as rtgvalidategraph, rtgapplylivegraphchanges, rtgstageknowledgechanges, rtgapplymigrationcutover, rtgexecutequery, snapshots, or ledger replay, especially for schema evolution, validation recovery…

volantlabs/vellis · 87 tokens

documentation-sync

Keep human documentation synchronized with canonical SysML v2 models, generated component and application views, implementation realizations, repository tooling, skills, and workflow rules. Use after model, code, generated-view, README, AGENTS, skill, command, or public behavior changes.

volantlabs/vellis · 57 tokens

python-component-implementation

Implement, review, or revise Python realizations of components from accepted textual SysML v2 contracts and generated views. Use when creating or aligning Python protocols, component classes, reference implementations, implementation bindings, or black-box contract tests.

volantlabs/vellis · 51 tokens

sysml-reference

Ground textual SysML v2 and KerML authoring, review, language-semantics decisions, construct selection, and official-validator diagnosis in the repository's checksum-pinned specification corpus. Use for questions or changes involving parts, references, features, actions, bindings, successions, flows, ports…

volantlabs/vellis · 92 tokens

model-hygiene-review

Audit accepted SysML v2 component or application models against implementations, protocols, consumers, tests, generated evidence, predecessor contracts, and git history before proposing remediation. Use for periodic model hygiene, suspected model/implementation drift, migration-loss reviews, verification-closure…

volantlabs/vellis · 72 tokens