Merge pull request #20 from swarmproof/release/v0.2.0

chore(release): v0.2.0

Author: MarouaBoud

.github/workflows/publish.yml

@@ -0,0 +1,40 @@
+name: publish
+
+# Publishes to PyPI when a GitHub Release is published, via PyPI Trusted Publishing
+# (OIDC — no API token/secret). One-time setup: register this repo + workflow as a
+# trusted publisher on PyPI (https://pypi.org/manage/account/publishing/). Until then
+# the build job still runs and attaches artifacts to the release.
+on:
+  release:
+    types: [published]
+  workflow_dispatch:
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+      - run: python -m pip install --upgrade build twine
+      - run: python -m build
+      - run: python -m twine check dist/*
+      - uses: actions/upload-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+  publish-pypi:
+    needs: build
+    runs-on: ubuntu-latest
+    if: github.event_name == 'release'
+    environment: pypi
+    permissions:
+      id-token: write  # required for Trusted Publishing (OIDC)
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+      - uses: pypa/gh-action-pypi-publish@release/v1

CHANGELOG.md

@@ -0,0 +1,35 @@
+# Changelog
+
+All notable changes to costbomb are documented here. Format loosely follows
+[Keep a Changelog](https://keepachangelog.com/); versions follow SemVer.
+
+## [0.2.0] — 2026-09-07
+
+First public release — the standalone extract of denial-of-wallet fuzzing.
+
+### Added
+- **Cost meter (the oracle)** — sum over sources: model tokens (input/output/reasoning/
+  cache) + tool-call fees + recursive sub-agent spawn cost. Provider-agnostic price table.
+- **Attack library** — 10 cost-explosion classes: `retry-loop`, `tool-storm`,
+  `context-bomb`, `recursion`, `clarification-trap`, `reasoning-inflation`,
+  `model-escalation`, `cache-bust`, `tool-cost-asymmetry`, `retrieval-amplification`,
+  each with honest capability gating.
+- **Fuzz engine** — evolutionary search, power schedule, p95-over-k fitness, surrogate
+  pre-ranking, and a hard own-budget cap (never runs away).
+- **CI gate** — baseline + regression detection with price-drift separation.
+- **Targets** — Fake / Python / HTTP / Mockworld / Persona, behind one `Target` seam.
+- **Proxy meter** — zero-instrumentation metering of a real agent via a base_url swap.
+- **Richer cost model** — downstream tool-cost / blast radius, wall-clock / infra cost,
+  and a duplicate-effect (exactly-once) cross-check backed by the real `exactly_once`
+  library.
+- **Validation** — per-biller reconciliation (`costbomb.validation`) + full-agent LLM
+  and Stripe test-mode harnesses; report labels modeled vs invoice-backed slices.
+- **CLI** — `costbomb run` / `baseline` / `proxy` / `attacks` / `price`, with
+  `findings.json` + OTel GenAI-profile export.
+
+### Known limitations
+- The meter's ≤1% accuracy (NFR-8) is **not yet validated against real invoices**
+  (`0/20` invoice-grounded fixtures) — see `docs/VALIDATION.md`. Treat as a
+  well-engineered tool whose oracle is arithmetic-checked but not bill-proven.
+
+[0.2.0]: https://github.com/swarmproof/costbomb/releases/tag/v0.2.0

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "costbomb"
-version = "0.2.0.dev0"
+version = "0.2.0"
 description = "Denial-of-wallet fuzzing for agent systems — find the inputs that make your agent spend $500 to answer a $0.05 question, and gate CI on spend regressions."
 readme = "README.md"
 requires-python = ">=3.11"

src/costbomb/__init__.py

@@ -23,7 +23,7 @@
 from costbomb.targets.base import Target, TargetContext
 from costbomb.targets.fake import FakeTarget
 
-__version__ = "0.2.0.dev0"
+__version__ = "0.2.0"
 
 __all__ = [
     "AttackClass",