You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: CHANGELOG.md
+28-1Lines changed: 28 additions & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -8,6 +8,31 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
8
8
9
9
Nothing yet.
10
10
11
+
## [0.3.0] — 2026-08-05
12
+
13
+
External key custody. `foc-cli` can now hold a *reference* to a key kept in a secret manager instead of the key itself, closing the gap that left MCP and CI with no option but a key in the config file. Contains one behaviour change to `wallet init` (hence the minor bump).
14
+
15
+
### Added
16
+
17
+
-`wallet init --keyRef <provider>:<reference>` — a third custody mode alongside a raw key and a Foundry keystore. The config stores only the pointer; the key is fetched into memory per command and never written to disk. Nothing prompts, so unlike keystore mode this works under the MCP server and CI. `--keyProject` scopes the reference; omitted, the provider picks its own default. `clawdi` is the first provider. ([#33])
18
+
-`wallet balance` now reports `keySource` (`keyRef` / `keystore` / `privateKey` / `none`), so a vault-backed setup is verifiable at a glance — the address proves which key signed, this proves where it came from. Never the key. ([#33])
19
+
- Preflight checks on every wallet-touching command, covering all three custody modes: `WALLET_NOT_CONFIGURED`, `MALFORMED_KEY_REF`, `KEY_REF_PROVIDER_MISSING`, `KEYSTORE_INTERACTIVE_ONLY` and `KEYSTORE_TOOL_MISSING` now arrive as typed errors with actionable CTAs, instead of escaping as an untyped throw from inside key resolution or from `cast`. The preflight does not resolve the key — that needs an authenticated provider and a round trip, and belongs at use time. A missing provider or missing Foundry is reported as `retryable` and deliberately carries **no** command: the wallet is intact and the fix lies outside foc-cli, so the only thing worth suggesting would have been one that discards a working configuration. ([#33])
20
+
-`wallet init --force`. ([#33])
21
+
22
+
### Changed
23
+
24
+
- **`wallet init` no longer silently replaces a configured wallet.** An explicit method used to overwrite whatever was configured, discarding a key that may have been the only copy. It now names what is at stake — the derived address for a private key, the path for a keystore, the reference for a key reference, never the key itself — and asks on a terminal, or fails with `WALLET_ALREADY_CONFIGURED` and a `--force` CTA in agent mode. The consequence stated is the one that actually applies: only a stored private key is destroyed by the swap, while a keystore file stays on disk and a vault key stays in the vault. Re-running the *same* method with the same value replaces nothing, and neither does adding or changing `--keyProject` on a configured reference — both are never blocked. That CTA replays the caller's own options minus the secret: a `--privateKey` passed on the refused invocation comes back as `0x...`, never the key. Automation that re-runs `wallet init --auto` expecting a fresh key must now pass `--force`. ([#33])
25
+
- Call-to-action guidance only offers a key-reference method when that provider's CLI is actually installed on the machine — suggesting a tool the caller does not have is a dead end. The reference docs still describe every provider. ([#33])
26
+
- Key resolution accepts a `0x` + 64 hex value only when it stands on its own, and refuses output holding more than one. A loose match was the dangerous case: every 32-byte value is a valid secp256k1 key, so the leading 64 hex digits of a longer blob would have been accepted and signed with — as a different address — rather than failing. ([#33])
27
+
- Windows: an npm-installed provider helper is a `.cmd`, which is a script rather than an executable and cannot be launched directly (Node has refused to since the fix for CVE-2024-27980). Those are now run through `cmd.exe`, with references restricted to a character set the shell treats literally so a tampered config still cannot become command execution. Previously the PATH probe found the helper, the preflight passed, and the launch failed with `EINVAL` — reported as "not logged in / key missing / wrong project", none of which was true. ([#33])
28
+
- A reference that is itself a private key (`clawdi:0x…` — the `--privateKey` mix-up with the provider prefix included) is refused at init and by the preflight rather than stored, and every message or result that quotes a reference redacts key-like runs first. A key is pure hex, so the character allowlist alone waved it through — stored under a field documented as safe to display, sent to the provider as a lookup name, and echoed verbatim into the use-time error envelope bound for the MCP result and the logs. ([#33])
29
+
30
+
### Documentation
31
+
32
+
-`references/key-injection.md` — identification table first (most of the time the answer is "already set up, run normally"), then setup, providers, what the mode does and does not protect, and the error catalog. ([#33])
33
+
-`references/integrations/clawdi-vault.md` — the Clawdi recipe, including the per-project scoping that most often bites. ([#33])
34
+
-`references/keystore-setup.md` now points at the key-reference mode as the automation-safe alternative it previously had no answer for. ([#33])
35
+
11
36
## [0.2.0] — 2026-07-23
12
37
13
38
Agent-hardening release ([#30]), driven by a 609-invocation live smoke campaign on Calibration and a keystore field test. Contains one breaking change (hence the minor bump).
@@ -106,7 +131,8 @@ Initial public release.
106
131
- MCP server mode and the two agent skills (`foc-cli`, `foc-docs`).
0 commit comments