diff --git a/.gitignore b/.gitignore
index 52c854d..8e63ef3 100644
--- a/.gitignore
+++ b/.gitignore
@@ -10,7 +10,6 @@ dist/
 build/
 *.egg-info/
 .tool-semantics/
-.agentabi/
 .env
 .DS_Store
 uv.lock
diff --git a/ROADMAP.md b/ROADMAP.md
index 40bc53c..651f0b5 100644
--- a/ROADMAP.md
+++ b/ROADMAP.md
@@ -10,35 +10,40 @@
 - [x] Publish repository
 
 ## Milestone 1 — Live MCP capture
-- Connect to local and remote MCP servers
-- Discover tools, prompts, and resources
-- Normalize metadata into a stable snapshot
-- Redact secrets and unstable fields
+- [ ] Connect to local and remote MCP servers (`capture --mcp`) — [#11](https://github.com/askmy-stack/tool-semantics/issues/11)
+- [ ] Discover tools, prompts, and resources
+- [ ] Normalize metadata into a stable snapshot
+- [ ] Redact secrets and unstable fields
 
 ## Milestone 2 — Compatibility engine
-- Required-parameter, type, enum, default, and output changes
-- Description-change warnings
-- Risk-level escalation detection
-- JSON and Markdown reports
-- CI-compatible exit codes
+- [x] Required-parameter, type, and enum changes
+- [x] Description-change warnings
+- [x] Risk-level escalation detection
+- [x] JSON and Markdown reports
+- [x] CI-compatible exit codes
+- [ ] Default-value and `output_schema` change codes — [#8](https://github.com/askmy-stack/tool-semantics/issues/8), [#15](https://github.com/askmy-stack/tool-semantics/issues/15)
+- [ ] Snapshot version validation on read — [#7](https://github.com/askmy-stack/tool-semantics/issues/7)
+- [ ] Ignore-config + verbose logging for CI adoption — [#9](https://github.com/askmy-stack/tool-semantics/issues/9), [#10](https://github.com/askmy-stack/tool-semantics/issues/10)
 
 ## Milestone 3 — Behavioral contracts
-- Positive, negative, and ambiguous intents
-- Side-effect and confirmation expectations
-- Human-reviewed probe generation
+- [ ] Positive, negative, and ambiguous intents — [#12](https://github.com/askmy-stack/tool-semantics/issues/12)
+- [ ] Side-effect and confirmation expectations
+- [ ] Human-reviewed probe generation
 
 ## Milestone 4 — Model matrix
-- Provider-neutral runner
-- Tool-selection and argument-validity metrics
-- Repeated trials and stability scoring
+- [ ] Provider-neutral runner
+- [ ] Tool-selection and argument-validity metrics
+- [ ] Repeated trials and stability scoring
+- [ ] Large-manifest benchmarks — [#14](https://github.com/askmy-stack/tool-semantics/issues/14)
 
 ## Milestone 5 — Pull-request reporting
-- Baseline versus candidate comparison
-- PR comments and detailed artifacts
-- Release-policy enforcement
+- [ ] Baseline versus candidate comparison
+- [ ] PR comments and detailed artifacts — [#13](https://github.com/askmy-stack/tool-semantics/issues/13)
+- [ ] Release-policy enforcement
+- [ ] PyPI publish / trusted publishing workflow
 
 ## Milestone 6 — Migration adapters
-- Tool aliases
-- Argument and enum translation
-- Output wrappers
-- MCP compatibility proxy
+- [ ] Tool aliases / rename detection
+- [ ] Argument and enum translation
+- [ ] Output wrappers
+- [ ] MCP compatibility proxy