Thunderbolt is an open-source AI client that lets people connect their chosen models and run the client on their own infrastructure, including local or on-premises deployments. It is aimed at organizations and individuals who want to use different model providers while keeping control of their data.
Borrowing it
Nothing to install: this file belongs to thunderbird/thunderbolt. 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/thunderbird/thunderbolt/main/.agents/skills/thunder-update-model/SKILL.mdgit clone --depth 1 https://github.com/thunderbird/thunderboltWrote 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/skills/thunderbird/thunderbolt/thunder-update-model)<a href="https://agentmods.dev/skills/thunderbird/thunderbolt/thunder-update-model"><img src="https://agentmods.dev/badge/skills/thunderbird/thunderbolt/thunder-update-model/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/skills/thunderbird/thunderbolt/thunder-update-model"><img src="https://agentmods.dev/badge/skills/thunderbird/thunderbolt/thunder-update-model.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 | $0.00042 | $0.03442 |
| Opus 5.5 | $0.00017 | $0.01377 |
| Sonnet 5.5 | $0.00008 | $0.00688 |
| Haiku 4.5 | $0.00004 | $0.00344 |
Grade A, and why
thunder-update-model scanned grade A with 1 finding 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 21d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
set -o pipefail; curl -fsS https://inference.tinfoil.sh/v1/models | jq -r '.data[] | select(.type=="chat") | "\(.id)\tdeprecated=\(.deprecated // false)\tdep_date=\(.deprecationDate // "-")\tctx=\(.context_window)\tmm=\( How it starts
The opening of the file, as written. The whole thing — 229 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Updating a built-in model
Swap one built-in model for its replacement, from a fresh worktree to a verified branch ready for
/thunderpush. The questions need the developer, so run this in the developer-facing session, not
in a subagent.
Read first
references/swap-checklist.mdin this folder. It is the implementer's contract; skim it now so the questions below collect everything it needs.docs/internals/architecture/managed-inference.md(routes, price rows, legacy slugs) anddocs/internals/architecture/reconciled-defaults.md#changing-a-default(version bump, snapshot test, one-way retirement).- The catalog itself,
shared/defaults/models.ts. Where a doc and the code disagree, the code wins.
Argument
Up to four space-separated tokens, in this order: the leaving model (slug or export constant,
never a display name), the provider that serves the replacement, the incoming model id, and a
Linear ticket (THU-nnn). Each token answers its question below; ask only the questions that
remain. Claude Code passes them as $ARGUMENTS; other harnesses pass the text typed after the
skill name.
Harness notes
Three actions differ by harness; every "ask" in the stages means the first row.
| Action | Claude Code | Any other harness |
|---|---|---|
| Ask the developer | AskUserQuestion. | Print the options as a numbered list in chat, after the stage's table if it has one, then wait for the developer's reply before continuing. Keep the question text as written. |
| Enter the worktree | EnterWorktree({ path }), loaded with ToolSearch (select:EnterWorktree). |
Prefix every later shell command with cd <absolute worktree path> && and use absolute paths. |
| Delegate the implementation | Agent tool. Pick the model and the effort for this swap from what the stages revealed: how much the catalog, the backend and the compatibility layer change, and how much of it the tests cover. Any model the harness offers is eligible, from the cheapest to the strongest, and effort moves on the same judgement. State the choice and the reason in one line before delegating. No isolation: "worktree". |
The same judgement, with the models that harness offers: a subagent where one exists, otherwise implement it yourself through the checklist. |
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 21d ago Changed · +3 lines 056c34c03751
- 22d ago First seen · 226 lines · 42 tokens per session scan A d8f95606eae7
thunder-update-model is a skill published in the GitHub repository thunderbird/thunderbolt (4,816 stars, last pushed today), licensed MPL-2.0. It adds 42 tokens to every session and 3,442 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-08.
Other skills, from other repositories
ci-commit
Only use when the user explicitly invokes this skill by name.
git-workflow
Git operations: commits, branches, PRs, and conflict resolution.
map-diff
Analyze the impact of staged git changes against the committed Cortex function knowledge graph — shows which downstream functions are affected before you commit.
git-analysis
Answer questions from repository history, including what changed, when, by whom, why a line exists, which commits touched an area, where a regression entered, and how branches relate. Use for blame, bisect-style narrowing, changelog drafting, and release diffs.
repository-writeback
Workflow for semantic-layer pull requests (editDbtProject), covering when to ask first, descriptions versus AI hints, impact and value checks, splitting work into pull requests, and the content migration after a merge. Load it before changing the dbt project and when a merged pull request needs follow-up.
lakefs-versioning
Use when lakeFS data versioning — Git-like branching for data lakes, atomic commits, time travel, CI/CD. Use when working with lakefs versioning.