feat: implement JIT Tools RAG injection with SubAgent discovery and cache

[dahai9]

Summary

Describe this PR in 2-5 bullets:

• Base branch target (master for all contributions): master
• Problem: The original cherry-pick for fcc5de41 had heavy merge conflicts with the newly refactored master branch.
• Why it matters: Tool RAG is critical for injecting dynamically discovered subagents to avoid context limits.
• What changed: Re-implemented the JIT Tools RAG integration from scratch atop the modern origin/master. Added fields to schema.rs, integrated initialization into start_channels(), and modified run_tool_call_loop parameters to support subset filtering via ToolRagIndex.
• What did not change (scope boundary): Did not change the core RAG chunking logic inside tool_rag.rs.

Label Snapshot (required)

• Risk label (risk: low|medium|high): risk: medium
• Size label (size: XS|S|M|L|XL, auto-managed/read-only):
• Scope labels (core|agent|channel|config|cron|daemon|doctor|gateway|health|heartbeat|integration|memory|observability|onboard|provider|runtime|security|service|skillforge|skills|tool|tunnel|docs|dependencies|ci|tests|scripts|dev, comma-separated): config, tool, agent, channel
• Module labels (<module>: <component>, for example channel: telegram, provider: kimi, tool: shell): tool: rag
• Contributor tier label (trusted contributor|experienced contributor|principal contributor|distinguished contributor, auto-managed/read-only; author merged PRs >=5/10/20/50):
• If any auto-label is incorrect, note requested correction:

Change Metadata

• Change type (bug|feature|refactor|docs|security|chore): feature
• Primary scope (runtime|provider|channel|memory|security|ci|docs|multi): multi

Linked Issue

• Closes #
• Related #
• Depends on # (if stacked)
• Supersedes # (if replacing older PR)

Supersede Attribution (required when Supersedes # is used)

• Superseded PRs + authors (#<pr> by @<author>, one per line):
• Integrated scope by source PR (what was materially carried forward):
• Co-authored-by trailers added for materially incorporated contributors? (Yes/No)
• If No, explain why (for example: inspiration-only, no direct code/design carry-over):
• Trailer format check (separate lines, no escaped \n): (Pass/Fail)

Validation Evidence (required)

Commands and result summary:

cargo fmt --all -- --check
cargo clippy --all-targets -- -D warnings
cargo test

• Evidence provided (test/log/trace/screenshot/perf): Standard test pipeline executed locally successfully.
• If any command is intentionally skipped, explain why:

Security Impact (required)

• New permissions/capabilities? (Yes/No): No
• New external network calls? (Yes/No): No
• Secrets/tokens handling changed? (Yes/No): No
• File system access scope changed? (Yes/No): No
• If any Yes, describe risk and mitigation:

Privacy and Data Hygiene (required)

• Data-hygiene status (pass|needs-follow-up): pass
• Redaction/anonymization notes:
• Neutral wording confirmation (use ZeroClaw/project-native labels if identity-like wording is needed): pass

Compatibility / Migration

• Backward compatible? (Yes/No): Yes
• Config/env changes? (Yes/No): Yes
• Migration needed? (Yes/No): No
• If yes, exact upgrade steps:

i18n Follow-Through (required when docs or user-facing wording changes)

• i18n follow-through triggered? (Yes/No): No
• If Yes, locale navigation parity updated in README*, docs/README*, and docs/SUMMARY.md for supported locales (en, zh-CN, ja, ru, fr, vi)? (Yes/No)
• If Yes, localized runtime-contract docs updated where equivalents exist (minimum for fr/vi: commands-reference, config-reference, troubleshooting)? (Yes/No/N.A.)
• If Yes, Vietnamese canonical docs under docs/i18n/vi/** synced and compatibility shims under docs/*.vi.md validated? (Yes/No/N.A.)
• If any No/N.A., link follow-up issue/PR and explain scope decision:

Human Verification (required)

What was personally validated beyond CI:

• Verified scenarios: Cargo checks for pipeline argument mismatches. Tests running with empty RAG indices via default configs.
• Edge cases checked: Addressed multimodal loops, interactive tools loop, and cli tests default None threading.
• What was not verified:

Side Effects / Blast Radius (required)

• Affected subsystems/workflows: Tools Registry, Agent Execution Loop
• Potential unintended effects: Tools failure if dynamically filtering removes essential base capabilities.
• Guardrails/monitoring for early detection: RAG relies on semantic thresholds.

Agent Collaboration Notes (recommended)

• Agent tools used (if any):
• Workflow/plan summary (if any):
• Verification focus:
• Confirmation: naming + architecture boundaries followed (AGENTS.md + CONTRIBUTING.md):

Rollback Plan (required)

• Fast rollback command/path: Revert commit
• Feature flags or config toggles (if any): disable tools_injection_mode = "rag", set back to "full"
• Observable failure symptoms:

Risks and Mitigations

List real risks in this PR (or write None).

• Risk: High retrieval threshold omits essential tools during long multi-turn execution.
  • Mitigation: tools_rag_threshold and tools_rag_core_set exist as config variables to manually bypass semantic failure loops.

[theonlyhennygod]

Closing as stale — this PR has had no activity for over 2 weeks and is significantly behind master. If you'd like to continue this work, feel free to reopen with a fresh rebase against current master. Thanks for the contribution!