interceptors

interceptors is an agent for coding agents from Bitterbot-AI/bitterbot-desktop. It costs 0 tokens per session (2,931 once invoked), scanned A, original, MIT.

Runtime checks that inspect an agent's proposed tool call before it runs and can change it, add context, require a prerequisite, or block it. They turn written behavior instructions into enforceable rules.

In plain words
What is it for?
Use them to enforce prerequisites and guardrails, inspect agent state, modify tool actions, inject relevant context, or stop calls that violate configured behavior.
Why use it?
They reduce cases where an AI agent ignores guidance about citations, channel etiquette, memory selection, or confidence because the rules are only written as prose.

Agent

About the project

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.

Bitterbot-AI/bitterbot-desktop · 2,460 stars · on GitHub

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 agents/bitterbot-ai/bitterbot-desktop/interceptors
Clone the repo
git clone --depth 1 https://github.com/Bitterbot-AI/bitterbot-desktop

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 interceptors

README.md
[![agentmods](https://agentmods.dev/badge/agents/bitterbot-ai/bitterbot-desktop/interceptors.svg)](https://agentmods.dev/agents/bitterbot-ai/bitterbot-desktop/interceptors)
Your own site
<a href="https://agentmods.dev/agents/bitterbot-ai/bitterbot-desktop/interceptors"><img src="https://agentmods.dev/badge/agents/bitterbot-ai/bitterbot-desktop/interceptors.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,931 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 $0.00000 $0.02931
Opus 5 $0.00000 $0.01465
Sonnet 5 $0.00000 $0.00586
Haiku 4.5 $0.00000 $0.00293

Measured 5d ago against content hash ea16a1fbf356, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

interceptors 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 5d 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.

docs/agents/interceptors.md · 241 lines

How it starts

The opening of the file, as written. The whole thing — 241 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Pre-Action Interceptors (PLAN-20)

Executable skill interceptors turn passive markdown skills into deterministic behavioral guardrails. When the agent proposes a tool call, registered interceptors inspect the action, read the agent's current state (hormonal, GCCRF, channel, recent turns), and may modify, inject context into, require a prerequisite for, or block the call before it executes.

This is the most important change to the skill system since PLAN-15. Where SICA gave us a staging gate for skill mutations, PLAN-20 gives us deterministic enforcement of skill behavior at runtime.

Why this exists

Bitterbot ships dozens of SKILL.md files. Each one is a piece of prose telling the agent how to behave. The LLM may or may not follow that prose on any given turn. Concrete daily losses we observed:

  • Citation rate on factual claims is ~40% despite the explicit cite-sources skill being loaded
  • Group-channel etiquette degrades under high engagement; the agent over-talks
  • Memory-tool selection drifts; relationship-shaped questions hit vector search and miss
  • Calibrated confidence (matching tone to actual epistemic state) is best-effort

Each of these is the same shape: skill says "do X under Y condition" → LLM mostly remembers → fails sometimes. PLAN-20 fixes "mostly" with deterministic dispatch driven by a typed activation predicate.

The conceptual lineage is the HASP paper (arXiv:2605.17734, May 2026), which reports ~25-30% task-completion lifts from doing exactly this. Bitterbot's contribution is binding the activation predicate to the agent's neuromodulatory + epistemic state, which no other framework can do.

Architecture

                         ┌─────────────────────────────┐
                         │  Built-in interceptors      │
                         │  src/agents/skills/         │
                         │   builtin-interceptors/     │
                         └──────────────┬──────────────┘
                                        │ register at autoboot
                                        ▼
   Agent proposes a tool call ──► runBeforeToolCallHook ──► tool executes
                                        │
                                        ▼
                              ┌───────────────────────┐
                              │   InterceptorRunner   │
                              │  - priority sort      │
                              │  - latency budget     │
                              │  - per-episode caps   │
                              │  - 3-strikes disable  │
                              └──────────┬────────────┘
                                         │
                ┌────────────────────────┼─────────────────────────┐
                │                        │                         │
       shouldActivate(ctx)      intervene(ctx, action)       persist + emit
                │                        │                         │
                ▼                        ▼                         ▼
        StepContext built from   modify / inject /        InterventionRecord
        session-context-tracker  require_prereq / block   (Ed25519-signed)
        + hormonal + GCCRF       / noop                   → sqlite v14

Read the full file on GitHub · 241 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. 5d ago First seen · 241 lines · 0 tokens per session scan A ea16a1fbf356

Subscribe to this mod's changes

interceptors is an agent published in the GitHub repository Bitterbot-AI/bitterbot-desktop (2,460 stars, last pushed today), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,931 tokens. 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 agents, from other repositories

domain

How the engineering skills should consume this repo's domain documentation when exploring the codebase.

rohansx/agidb · 0 tokens

triage-labels

The skills speak in terms of five canonical triage roles. This file maps those roles to the actual label strings used in this repo's issue tracker.

rohansx/agidb · 0 tokens

issue-tracker

Issues and PRDs for this repo live as GitHub issues at github.com/agidb/agidb (to be created in phase 0). Use the gh CLI for all operations.

rohansx/agidb · 0 tokens

nextjs-expert

Next.js framework strategist. Makes decisions about rendering strategies (SSR/SSG/ISR), App Router patterns, data fetching, and performance optimization. Use when designing Next.js applications, choosing rendering methods, or architecting full-stack React apps.

armanzeroeight/fastagent-plugins · 53 tokens

refactoring-architect

Strategic refactoring agent specializing in code improvement decisions, architectural enhancements, and technical debt management. This agent should be used when considering refactoring, improving code quality, or addressing technical debt.

armanzeroeight/fastagent-plugins · 43 tokens

example-org-coordinator

Single entry point for the example-org engagement — board grooming on the example-board GitHub Project, status-report prep for the example-team mandant, and routing per rules/org/example-routing.md. Spawn for board sync, weekly status collection, or any example-org coordination that would dump too much raw output into…

bks-lab/open-bridge · 69 tokens