[V3] #233: Structured tool arguments and MCP invocation

[EvanProgramming]

Summary

Completes V3 issue #233 (R014–R018). It routes provider and MCP calls through the same validated JSON-object executor boundary, with typed results and structured receipts.

Changes

• Added strict object-schema validation for required/unknown fields, nested objects and arrays, primitive types, constraints, combinators, finite JSON values, and unsupported keywords.
• Replaced generic built-in schemas with useful file, shell, Git/GitHub, browser, graph, memory, discovery, calculation, and orchestration schemas.
• Passed MCP arguments directly into AgentRuntime.execute_call(); MCP error status now follows typed receipt success.
• Migrated built-in adapters and runtime observers to preserve structured arguments through execution, scope checks, receipts, effect verification, and acceptance evidence.
• Kept text APIs and legacy action parsing at compatibility edges. Legacy string-only custom executors remain text-only; dynamic tools receive declared object arguments and return ToolResult.
• Addressed all eight actionable Codex review threads with focused regressions, including Git flag bypass, structured artifact receipts, browser close forwarding, optional schema fields, MCP aliases, malformed legacy quoting, decimal multipleOf, and inherited capability ceilings.
• Regenerated the tool inventory and updated the related subagent acceptance fixture for structured receipt arguments. Related malformed-search coverage is linked to [enhancement] Recover from malformed search_files arguments #226.
• Based on merged [V3] Preserve capability gates, structured approvals, and narrowing permissions #261 security changes; the latest signed sync commit preserves the exact tree tested locally.

Validation

• PASS — 673-test suite on Python 3.12 on the exact same tree as the final sync commit.
• PASS — 123 focused tests on each of Python 3.12 and 3.13 for [V3] Migrate native tool arguments and MCP to one typed invocation boundary #233; [V3] Preserve capability gates, structured approvals, and narrowing permissions #261 focused tests pass separately on both versions.
• PASS — make check, make lint, make docs-check, Go TUI tests, and fixture-based subagent acceptance.
• PASS — independent review of final residual fixes found no remaining actionable issue; all 8 inline threads are resolved.
• PASS — local and GitHub GPG verification for final sync commit d7ae60d.
• PASS — GitHub OpenKyrozen CI and publish workflows on the final main-target head.
• SKIPPED — live provider calls; changes validate offline and do not require a provider. CodeRabbit automation is disabled for this base branch.

Compatibility and security

Existing text-facing APIs remain available through compatibility adapters. Native custom calls require object arguments matching their declared schema; legacy string-only executors are rejected at the native boundary. No execution retry, provider call, dependency, or merge is included.

This PR targets main and is ready to merge.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

• Configuration used: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: 63ee6dcb-b82c-416e-bc2d-051d0e858767

---

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

---

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-11T16:05:59.773259Z	a51b911	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a51b9114b8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".