OpenClaw merged PR #119306, "fix(commands): restore tool inventory for dynamic models", a P2 command reliability fix for sessions that use dynamically prepared provider models.
The user-facing symptom was simple: /tools could return a temporary inventory failure even when the session had valid tools available. The underlying issue was timing and ownership. The command tried to resolve the effective tool inventory before publishing or borrowing the prepared runtime model context needed by dynamic-provider sessions.
That made an informational command look unreliable in exactly the sessions where operators are most likely to need it: model-aware, dynamically configured environments with provider-specific tool surfaces.
Runtime Context First
The fix changes the /tools path to follow the Gateway owner boundary more carefully. OpenClaw now resolves the effective runtime model context asynchronously, then passes that prepared context into the synchronous tool-inventory projection.
That means the inventory resolver receives the same kind of model information the live session is actually using, including the runtime model and model API shape. The change is intentionally narrow: it repairs the inventory path without changing the command's output format or broadening access to tools.
For users, the result is that /tools should once again show available built-in and connected tools instead of a generic "try again" style failure.
Release QA Tightening
The PR also repaired a separate Beta 8 release-lane gate. OpenClaw's environment-variable budget expected 515 distinct production OPENCLAW_* names, while the current observed count had dropped to 513 after two names were removed. The ratchet now matches the repository checker so check:changed can pass on current main.
The Telegram release-QA scenario for the compact /tools command was tightened at the same time. It now uses the established model-aware live-turn timeout and asserts the full expected reply. That makes fallback inventory failures visible in release proof instead of letting the scenario accept an unexpected response.
User Impact
The practical impact is a cleaner operational loop. If a user asks OpenClaw what tools are available in a dynamic-provider session, the command should resolve against the prepared model context and return the actual inventory.
That matters for troubleshooting, onboarding, and command discovery. Tool availability is not just decoration; it tells users what the current assistant can do and whether connected tools were loaded correctly.
The release-gate repair is less visible but still important. A broken ratchet can block release validation even when production code is fine. Keeping that check aligned lets release automation catch real environment-surface growth without failing on already-removed names.
Evidence
PR #119306 reports a clean autoreview after the ratchet repair with confidence 0.98. Local focused proof covered 98 tests across commands-info.test.ts and live-scenario-timeouts.test.ts, plus a confirmed OPENCLAW_* count 513/513 and clean git diff --check.
The Blacksmith Testbox run passed full pnpm check:changed plus the same 98 focused tests. The AWS Crabbox/Mantis Telegram proof also passed the telegram-tools-compact-command scenario at the exact candidate SHA, with the report containing the expected redacted reply: available tools, profile, built-in and connected tools, and the /tools verbose hint.
For OpenClaw operators, this is a small command fix with outsized usefulness. The /tools command is back to being a reliable map of the active session instead of another thing to debug.