@codacy/verity-cli 0.34.1 → 0.34.2-experimental.312358d

package/CHANGELOG.md

@@ -64,6 +64,34 @@ installs are untouched.
   which is the ordinary case for someone whose first Verity install *is* the
   plugin.
 
+## [0.34.2] — 2026-10-12
+
+**Commands you type keep their Claude Code session, and the version warning
+shows once a day.**
+
+### 🐛 Fixes
+
+- **Typed commands keep the session.** `verity review`, `verity reflect`,
+  `verity ignore` and the other commands that an agent types now read the
+  session id that Claude Code sets (`CLAUDE_CODE_SESSION_ID`). Their runs join
+  the session's task, instead of starting a new task for every call.
+- **`verity review --files` and `--changed` stay inside the project.** A path
+  outside the project is refused with a warning that names it, and nothing is
+  sent. A path written through a linked directory is accepted.
+- **The version mismatch warning shows once a day.** When the global `verity`
+  and the plugin's hooks run different versions, the warning shows at the first
+  session of the day, and again when the versions change. The agent still gets
+  the facts at every session start.
+- **`verity prove generate` uses the exact lines of the function.** It finds them
+  in the code map, as the push check does. A name that is not in the file stops
+  before any request and lists the names that the file has.
+- **Formal verification's records no longer replace your review history.**
+  After a formal check, the pending items and the last verdict come from the
+  last review, not from the check's own records.
+- **A push with formal verification records a summary of the check.** The
+  summary holds names, line numbers and counts only, never source code.
+- Includes security fixes.
+
 ## [0.34.1] — 2026-10-08
 
 **Fixes from user feedback: large changes get reviewed again, and `verity init`

package/README.md

@@ -176,6 +176,13 @@ No re-setup needed: your token, Standard, and run history all carry over.
 | `verity reflect --kind decision` | Specify the node kind |
 | `verity memory untrack` | Stop committing `.verity/memory/` — the graph stays on disk and leaves your diffs |
 | `verity memory track` | Commit the graph on purpose, and record that so nothing offers to change it |
+| **Formal verification** | |
+| `verity config formal push` | Opt this checkout in: at push, one risky changed function may get a Lean model, checked and replayed (see below) |
+| `verity config formal off` | Opt out again |
+| `verity lean status` / `install` / `remove` | The managed Lean 4 toolchain the models are checked with (≈800 MB download, 2.7 GB on disk) |
+| `verity prove candidates [paths]` | Rank the riskiest code from git history and shape signals, no model |
+| `verity prove auto [files]` | Run the push-time phase by hand on the commits ahead of upstream |
+| `verity prove waive <file> --reason "…"` | Dismiss a reproduced counterexample for these exact bytes; it returns if the file changes |
 | **Feedback** | |
 | `verity feedback message "text"` | Send general feedback |
 | `verity feedback message -` | Same, reading the message from stdin |
@@ -196,6 +203,37 @@ No re-setup needed: your token, Standard, and run history all carry over.
 | `/verity-memory` | Browse the knowledge graph |
 | `/verity-insights` | Quality metrics + Standard evolution |
 | `/verity-reflect` | Capture learnings after a task |
+| `/verity-prove <worry>` | Formal verification on demand: pick a function together, model it in Lean, replay any counterexample on the real code |
+
+## Formal verification (opt-in)
+
+At `git push`, beside the review, Verity can build a **small Lean 4 model of one
+risky function in the change**, search every interleaving of its actions up to a
+bound, and, when a property breaks, **replay the failing trace on the real code**
+in a disposable container. Only a counterexample that reproduces on the real
+code can block the push; a model's opinion alone never does.
+
+**Most pushes get no formal check.** The phase runs only when the push touches a
+function that looks worth it: history says the file has been fixed before, the
+text shows a risky shape (check-then-act, retries, money, auth, a transaction, a
+cache), and a cheap ranking call scores it above the bar. A push of docs, tests
+or plain code stops in under a second with a named reason. Counterexamples can
+only be replayed for **TypeScript, JavaScript and Python**; other languages stop
+before any model is paid for.
+
+```bash
+verity config formal push     # opt in (verity init asks on an interactive terminal)
+verity lean install           # the toolchain, once per machine
+verity prove auto             # try it by hand on what is ahead of upstream
+```
+
+The service keeps a structured record of each phase (the function, the model's
+name, each promise's id and the line number or rule it rests on, each
+counterexample's trace of action names and its verdict) and never the code, the
+model, the harness or any prose about them. The full account — what
+gets selected, the five structural rules, what a replay proves, costs and caps,
+and what the dashboard shows — is at
+https://verity.md/docs/formal.
 
 ## How it works