tpu_performance_autoresearch_wiki: Skill for Claude Code

.claude/skills/ingest-source/SKILL.md

ingest-source is a skill for Claude Code from vlasenkoalexey/tpu_performance_autoresearch_wiki. It costs 173 tokens per session (2,600 once invoked), scanned A, original, MIT.

A controlled workflow for adding a paper, article, talk, blog post, or documentation page to a project wiki. It reads the source, discusses which points matter, writes a source page, and updates related connections.

In plain words
What is it for?
It is used to ingest a local source file or fetched web page into the wiki, summarize its relevant claims, file the result, and connect it to related concepts, models, or codebases.
Why use it?
It keeps imported research organized and prevents important ideas from entering the wiki without deciding how they should be emphasized or linked to existing work.

Skill for Claude Code

Written for Claude Code: installed under .claude/. Also seen: reads .claude/ paths; names the AskUserQuestion tool.

This is vlasenkoalexey/tpu_performance_autoresearch_wiki's own configuration. It tells Claude Code how to work on tpu_performance_autoresearch_wiki itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything tpu_performance_autoresearch_wiki configures →

Reuse

Borrowing it

Nothing to install: this file belongs to vlasenkoalexey/tpu_performance_autoresearch_wiki. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/vlasenkoalexey/tpu_performance_autoresearch_wiki/main/.claude/skills/ingest-source/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/vlasenkoalexey/tpu_performance_autoresearch_wiki

Made for: Claude Code.

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-source

README.md
[![agentmods](https://agentmods.dev/badge/skills/vlasenkoalexey/tpu_performance_autoresearch_wiki/ingest-source/github.svg)](https://agentmods.dev/skills/vlasenkoalexey/tpu_performance_autoresearch_wiki/ingest-source)
Your own site
<a href="https://agentmods.dev/skills/vlasenkoalexey/tpu_performance_autoresearch_wiki/ingest-source"><img src="https://agentmods.dev/badge/skills/vlasenkoalexey/tpu_performance_autoresearch_wiki/ingest-source/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.

agentmods 80×15 button for ingest-source

Your own site · 80×15
<a href="https://agentmods.dev/skills/vlasenkoalexey/tpu_performance_autoresearch_wiki/ingest-source"><img src="https://agentmods.dev/badge/skills/vlasenkoalexey/tpu_performance_autoresearch_wiki/ingest-source.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 173 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,600 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector warn 7 Sept 2026
SkillSpector: 2 findings, up to medium

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

  • medium Agent Snooping · line 210
    Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
    Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
  • medium Agent Snooping · line 211
    Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
    Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
How audits are shown
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 5.1 $0.00173 $0.02600
Opus 5 $0.00086 $0.01300
Sonnet 5 $0.00035 $0.00520
Haiku 4.5 $0.00017 $0.00260

Measured 12d ago against content hash 82f9eb00e0d4, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade A, and why

ingest-source 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 12d 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-source/SKILL.md · 212 lines

How it starts

The opening of the file, as written. The whole thing — 212 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/ingest-source — ingest a paper/article/talk/docs into the wiki

Encodes SCHEMA's INGEST-SOURCE operation. The discipline matters: ingest is the front door for the wiki's hypothesis-generation pipeline, and getting emphasis wrong (or skipping connection updates) propagates downstream into bad hypothesis proposals.

When to invoke

  • User says "ingest this paper / blog post / talk / docs page" and provides a URL or raw/sources/ path.
  • A paper or talk just landed that touches TPU perf-relevant topics (Pallas kernels, sharding partitioners, FP8 training, attention kernels, scheduler/compiler passes, etc.).
  • An existing source needs re-ingestion because a meaningful update was published.
  • NOT for codebases — those go through INGEST-CODEBASE (separate skill, TBD).
  • NOT for per-experiment narrative — that's wiki/experiments/ and a different operations contract.

Cardinal rules

  • Discuss before writing. The human owns emphasis decisions. Do NOT write the source page until you've surfaced the key claims and confirmed which ones matter.
  • Update connections. Every concept, model, or codebase that the source touches gets its ## Sources updated and (where appropriate) prose extended. Skipping this step is the canonical failure — orphan source pages with no incoming links.
  • Generate hypothesis candidates. If the source suggests optimizations applicable to an active model, file them as wiki/hypotheses/<slug>.md with origin: <source-slug>. Add to the model page's ranked list.
  • Update index.md and log.md. Mandatory per SCHEMA's operations contract.
  • Never modify raw/. The source file is immutable. If the URL was given, save to raw/sources/<year>-<slug>.<ext> first.

Step 1 — Acquire the source

The user supplies one of:

  • A raw/sources/<year>-<slug>.<ext> path (already ingested locally)
  • A URL (fetch + save)
  • A direct file paste

If URL: fetch and save to raw/sources/<year>-<slug>.<ext>. Choose <slug> to be short, lowercase, hyphen-separated, descriptive. Choose <year> from the source's publication date.

Read the full file on GitHub · 212 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. 12d ago First seen · 212 lines · 173 tokens per session scan A 82f9eb00e0d4

Subscribe to this mod's changes

ingest-source is a skill published in the GitHub repository vlasenkoalexey/tpu_performance_autoresearch_wiki (55 stars, last pushed 8d ago), licensed MIT. It adds 173 tokens to every session and 2,600 once invoked, about $0.0009 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.