Bitterbot is a local-first personal AI agent that runs on a user’s devices, keeps persistent memories, performs tasks, and can exchange reusable skills with other agents. It is intended for people who want a personal assistant that remains available across conversations and activities. The catalogue entries provide instructions and agents for working with Bitterbot.
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.
npx skills add Bitterbot-AI/bitterbot-desktop --skill calibrate-claim-confidencegit clone --depth 1 https://github.com/Bitterbot-AI/bitterbot-desktopWrote 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/bitterbot-ai/bitterbot-desktop/calibrate-claim-confidence)<a href="https://agentmods.dev/skills/bitterbot-ai/bitterbot-desktop/calibrate-claim-confidence"><img src="https://agentmods.dev/badge/skills/bitterbot-ai/bitterbot-desktop/calibrate-claim-confidence/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/bitterbot-ai/bitterbot-desktop/calibrate-claim-confidence"><img src="https://agentmods.dev/badge/skills/bitterbot-ai/bitterbot-desktop/calibrate-claim-confidence.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00046 | $0.00357 |
| Opus 5 | $0.00023 | $0.00179 |
| Sonnet 5 | $0.00009 | $0.00071 |
| Haiku 4.5 | $0.00005 | $0.00036 |
Grade A, and why
calibrate-claim-confidence 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.
What it actually says
calibrate-claim-confidence
Bitterbot has an epistemic state (the GCCRF reward function) that quantifies how empowered the agent feels by its current knowledge: high empowerment means it has corroborated context, low means it's running on uncertain ground. When the agent is about to send a message containing confident absolutes ("definitely", "certainly", "always", "100%") but its empowerment is low and its certainty is dropping, this interceptor rewrites the message into hedged language.
This is the canonical example of state-binding: no other agent framework reads gccrf.empowerment to decide whether to hedge an outgoing statement.
What you'll see
When the agent is uncertain, you will see softer language: "likely" instead of "definitely", "typically" instead of "always", "it appears" instead of "obviously". When it is confident (high empowerment, rising certainty), absolutes are left intact.
Implementation
Built-in interceptor calibrate-claim-confidence:default lives in src/agents/skills/builtin-interceptors/calibrate-claim-confidence.ts. Fires up to 6 times per session.
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.
- 12d ago First seen · 27 lines · 46 tokens per session scan A 40c07318651e
calibrate-claim-confidence is a skill published in the GitHub repository Bitterbot-AI/bitterbot-desktop (2,462 stars, last pushed yesterday), licensed MIT. It adds 46 tokens to every session and 357 once invoked, about $0.0002 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.
Other skills, from other repositories
selfhost-emem-guard
Stand up an emem-guard verdict server, verify it against the conformance checks, and point any agent at it. Use when asked to self-host emem-guard, add a grounding gate to an agent on any model or framework, wire a checkpoint into Claude Code or Claude Enterprise or MCP, or run a signed allow/deny server for…
emem-field-tokens
Get a native-resolution raster field over an area from emem, or a field over time, as a signed, verifiable artifact rather than a set of per-cell scalars. Use when the user needs the actual grid of values over an area of interest (a world model input, an NDVI/band drape, change analysis over a scene window, exportable…
emem-a2a-collaboration
Join the agent-to-agent collaboration running on emem's signed ledger — find the standard, verify another agent's message offline (who wrote it, not just that it was stored), announce yourself, and hand facts to other agents as tokens. Use when the user wants agents to coordinate without a shared database or shared…
emem-find-similar
Given a place name or cell64, return the top-K most similar places on Earth by cosine similarity over the 128-D Tessera foundation embedding. Use when the user asks for analogues, look-alikes, or counterparts ("find cities like Bangalore", "where else looks like the Sundarbans", "show me places with a similar urban…
emem-locate-and-recall
Resolve a free-form place name to an emem cell64 and recall signed Earth-observation facts at that location. Use when the user asks about current weather, vegetation index, elevation, soil properties, or any other geospatial measurement at a named place ("what's the temperature in Bengaluru", "how high is Denali"…
emem-recall-polygon
Recall signed Earth-observation facts at every cell inside a user-supplied polygon. Use when the user asks about an extent rather than a point — "what's the average NDVI inside this watershed", "show me precipitation across the Western Ghats", "what's the elevation profile of this region". Accepts a polygon as [lng…