Skip to content
Back to search

temet-ai

MarketplaceWorks with Claude CodeGitHub

Claude Code plugins from temet-ai

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 20 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/temet-ai/ask-kit with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/temet-ai/ask-kit'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/temet-ai--ask-kit/badge.svg)](/listings/temet-ai--ask-kit)

HTML

<a href="/listings/temet-ai--ask-kit"><img src="/api/listings/temet-ai--ask-kit/badge.svg" alt="findsafeskills safety rating" /></a>

Install

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 12 lines
  1. {
  2. "name": "temet-ai",
  3. "owner": { "name": "Dawid du Toit", "url": "https://github.com/temet-ai" },
  4. "description": "Claude Code plugins from temet-ai",
  5. "plugins": [
  6. {
  7. "name": "ask-kit",
  8. "source": "./",
  9. "description": "One plain-language `ask` tool: a cheap model answers about your past sessions, your code and any kit you add, outside your agent's context."
  10. }
  11. ]
  12. }
README.md · README · 159 lines
  1. <h1 align="left"><a href="https://github.com/temet-ai/ask-kit"><img src="docs/assets/ask-kit-logo.png" alt="ask-kit — playful brown stacked wordmark" width="240"></a></h1>
  2. Most of a coding agent's context goes on finding things out: searching, opening files, following
  3. calls, checking what was tried before. All of it stays in the context and crowds out the work
  4. the agent was asked to do.
  5. ask-kit keeps your coding agent's context small by moving that work out of it. It gives the
  6. agent one tool, `ask`, that answers plain-language questions about the repository's code and
  7. searches past Claude Code, Codex, Copilot CLI, Gemini CLI and opencode sessions. A cheap investigator works it out in
  8. its own context, and the agent gets back a short answer with the evidence it's based on
  9. (`path:line`, or the agent, session and event).
  10. ```
  11. you: add a CSV exporter like the other ones
  12. agent → ask: how are the exporters built and registered, and which tests cover them?
  13. ask → Exporters implement Exporter in src/export/base.ts:12 and are registered in
  14. src/export/index.ts:30. JsonExporter (src/export/json.ts) is the closest model;
  15. tests are in test/export/*.test.ts.
  16. ```
  17. (This is an example of what an answer looks like.
  18. [Context, time and cost per ask](#context-time-and-cost-per-ask) has real numbers.)
  19. A real ask from the shell, about this repository's code, with a line per program the investigator
  20. runs and then a follow-up:
  21. ![ask-kit ask on this repository: a line per cell while the investigator works, the answer with path:line citations, then a --continue follow-up answered in 8 seconds](docs/user-guide/images/ask-from-shell.gif)
  22. `ask` is an [MCP](https://modelcontextprotocol.io) server, so any agent that takes MCP servers
  23. can use it: Claude Code, Codex, GitHub Copilot CLI, Gemini CLI, opencode.
  24. **[Read the guide](docs/user-guide/README.md)**: how to use it, use cases, models, tuning. Also [read it on temet.ai](https://temet.ai/docs/ask-kit/).
  25. ## How the investigator answers
  26. For each step of a question, the investigator writes a small JavaScript program and runs it in a
  27. REPL. The program does whatever the step needs: loops, filters, joins, counts, comparisons,
  28. parsing. It is written fresh for that question. A domain gives the programs access to the data.
  29. What a program prints is all that returns to the investigator's model. Code mode, recursive
  30. language models and Claude Code subagents each do part of this;
  31. [How ask-kit relates to them](docs/user-guide/code-mode.md) goes through each.
  32. ![Your agent sends one question; the investigator writes small programs for it in its own context, reaching code, sessions and your own domains, and returns a short answer with path:line evidence](docs/user-guide/images/how-an-ask-works.svg)
  33. Each thing `ask` can answer about is a **domain**. ask-kit comes with two, as a starting point:
  34. - **code**: the repository the agent is working in, cited as `path:line`.
  35. - **sessions**: your past agent sessions, read from the transcripts each agent already saves,
  36. where they are, with no proxy or database.
  37. [Add your own](docs/user-guide/domains.md) for anything else your agent keeps digging through.
  38. ## Set it up
  39. Paste one line into the agent you already use, on Linux, macOS or Windows:
  40. ```
  41. Read https://github.com/temet-ai/ask-kit/blob/main/AGENTS.md, then set up ask-kit for me.
  42. ```
  43. The agent installs ask-kit and asks you a few questions. The main one is which model does the
  44. digging:
  45. - an API key (cheapest and fastest),
  46. - a subscription you already pay for (Claude Code, Codex, Copilot),
  47. - or a local model.
  48. It shows what it will change before it changes anything. Restart your agents afterwards, and
  49. they call `ask` by themselves when they need to find something out.
  50. In Claude Code you can install it as a plugin instead. Its npm packages come with the install, and
  51. it needs Node.js 22.13 or newer:
  52. ```
  53. /plugin marketplace add temet-ai/ask-kit
  54. /plugin install ask-kit@temet-ai
  55. ```
  56. The plugin brings the `ask` tool, the skills, the tuner agent and two hooks. After the third
  57. lookup in a row that `ask` could have answered, a PostToolUse hook adds one line to Claude's
  58. context naming the investigator that fits, and a SubagentStart hook gives each subagent the short
  59. rule. Then ask Claude to set up
  60. ask-kit, and the ask-kit-setup skill asks you the same questions. The plugin covers Claude Code
  61. only; `ask-kit init` and `ask-kit setup` leave Claude Code alone while it is installed.
  62. In Gemini CLI, `gemini extensions install https://github.com/temet-ai/ask-kit` adds the `ask`
  63. tool, run through `npx @temet-ai/ask-kit`. Run `npx @temet-ai/ask-kit setup` once to choose the
  64. model.
  65. Without an agent: `npm install -g @temet-ai/ask-kit`, then `ask-kit init` and `ask-kit setup`.
  66. It needs Node.js 22.13 or newer, and git for the code domain's history and for shared kits. If npm answers `EACCES`, add
  67. `--prefix ~/.local` to the install command.
  68. [Set it up](docs/user-guide/setup.md) has the details.
  69. ## Context, time and cost per ask
  70. How much context one ask keeps out of the calling agent. These are medians from mike-ask, the
  71. proof of concept ask-kit grew from:
  72. ![Median per ask over 105 real asks to mike-ask: the investigator worked through 125,025 tokens, and about 445 went back to your agent; 36 seconds and $0.011](docs/user-guide/images/context-kept-out.svg)
  73. ask-kit itself, on 18 Nov 2026, on one machine with about 1,500 sessions on disk:
  74. | Question | Investigator | Time | Cost |
  75. |---|---|---|---|
  76. | What did a Codex session find and change | glm-5.3-flash (OpenRouter) | 16 s | $0.0022 |
  77. | Which sessions worked on X in the last week | glm-5.3-flash, domain picked by the router | 53 s | $0.0083 |
  78. | Follow-up on that (`continue: true`) | glm-5.3-flash | 17 s | $0.0035 |
  79. | Which Copilot session did a piece of work, under which ticket | Claude Code, haiku | 23 s | $0.0025 |
  80. | same | Codex (ChatGPT plan) | 36 s | in the plan |
  81. | same | Copilot CLI | 134 s | Copilot premium requests |
  82. | How does routing work in this repo | deepseek-v4.1-flash (code) | 110 s | $0.0033 |
  83. ## What's in this repository
  84. | Folder | What |
  85. |---|---|
  86. | `bin/`, `src/` | the `ask-kit` command, the MCP server, the investigator loop, the transcript readers |
  87. | `domains/` | the two built-in domains, `code` and `sessions`: instructions, scripts and the typed description of what they can reach |
  88. | `bundle/` | what `ask-kit init` installs into each agent: three skills, the `ask-kit-tuner` agent and the when-to-ask rule. They are templates, and `init` fills in this machine's paths. |
  89. | `docs/user-guide/` | the guide, also at temet.ai/docs/ask-kit |
  90. ## Status
  91. Early. Tested on Linux with real transcripts. Fixtures for all five transcript formats run in CI
  92. on Linux, macOS and Windows. The Gemini CLI reader and the opencode and Gemini install formats
  93. follow their docs and haven't yet been tried on a real install. Releases are on npm as
  94. [`@temet-ai/ask-kit`](https://www.npmjs.com/package/@temet-ai/ask-kit).
  95. ask-kit started as mike-ask, a proof of concept that answers from what
  96. [mike](https://github.com/temet-ai/mike) records of every Claude Code session. ask-kit is the
  97. general version. It needs no mike and no server.
  98. ## What it stores and sends
  99. It sends your question, and the excerpts the investigator reads, to the model you chose (and, for
  100. a slow ask, to the second backend you set with `escalate` or `hedge`). Keys, tokens and passwords
  101. in transcript excerpts are masked before the model sees them. There's no telemetry. Everything it keeps is under `~/.config/ask-kit/` and `~/.local/state/ask-kit/`,
  102. including the clipped prompts `ask-kit gaps` caches from your transcripts, a digest of each
  103. session `find-sessions` matched (title, first prompt, top words), a lookup counter per
  104. Claude Code session and a log of the nudges the plugin's hook gave (time, session id, kind of
  105. lookup; no commands or prompts) and, for the one in 20
  106. action answers that a model checks in the background, both answers. Each ask's journal row
  107. holds how many of the answer's citations resolved, and the ones that did not (a path and line,
  108. or a session and event); the check reads local files and transcripts only. With `journal.trajectories:
  109. true` (off by default) the journals also keep each cell's printed result and each answer, so they
  110. then hold excerpts of what was read.
  111. [What it stores and sends](docs/user-guide/privacy.md) has the full list.
  112. ## Contributing
  113. Bug reports and small fixes are welcome; see [CONTRIBUTING.md](CONTRIBUTING.md). Report security
  114. problems privately, as [SECURITY.md](SECURITY.md) says.
  115. ## Licence
  116. MIT, see [LICENSE](LICENSE). Third-party licences are in
  117. [THIRD-PARTY-NOTICES.md](THIRD-PARTY-NOTICES.md).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.