isAllCapsPlaceholder() classified ANY string of >=3 chars made only of A-Z and
underscore as an LLM placeholder, and stripEmptyOptionalArgs() then removed it
from the outbound CallTool arguments. Real business identifiers match that shape:
bank codes BKASH, NAGAD, ROCKET, VCB
currency codes BDT, VND, USD, PKR, MMK
enum values SUCCESS, PENDING, ALL
Impact is silent and severe: dropping an optional filter turns a filtered query
into an unfiltered one. The MCP server still answers 200 with the full result set,
so nothing errors — the model reports totals for the entire dataset as if they were
the filtered result, with full confidence.
Observed in production: asking "how many BKASH proxies are active" returned 6,610
(all banks) instead of 1,263. Three different bank codes returned byte-identical
results, which also drove the model into a retry loop until the loop guard fired.
Diagnosis was slow because the removal is invisible: the outbound log records
args_len BEFORE stripping, and stripEmptyOptionalArgs logs nothing at all — while
normalizeArgsForSchema right below it does emit mcp.tool.args.coerced. Confirmed
the argument never reaches the server by echoing the raw JSON-RPC arguments from
RequestContext[CallToolRequestParams] on the server side.
Changes:
- isAllCapsPlaceholder: require multi-word (contains "_") or membership in a known
single-word placeholder list. Keeps the original intent — the doc comment's own
examples (SHOULD_NOT_BE_HERE, DO_NOT_SEND, NOT_APPLICABLE) are all multi-word —
while letting business codes through.
- stripEmptyOptionalArgs: emit slog.Warn("mcp.tool.args.stripped") whenever an
argument is dropped, so this class of bug is one grep away instead of a
multi-hour investigation.
- Tests: guard both layers against regression.
Verified: go build ./... clean; go test ./internal/mcp/ passes.
Resolve the PR #8/#9/#10 stack on current dev, including Bitrix24 install callback hardening, migration renumbering, duplicate-domain fail-closed routing, UI textarea/mobile cleanup, and review hardening.
- Add GrantChecker interface with cache + event-bus invalidation for revalidating MCP tool grants at execution time (not just dispatch)
- Move toolGroups from package-level global to per-Registry field; Clone() gives each agent Loop isolated tool groups, eliminating cross-agent race condition
- BridgeTool.Execute rechecks grant validity before executing MCP methods; uses serverID field for cache key
- Wire MCPGrantChecker through ResolverDeps to MCP Manager and LoopConfig
- Add integration test for grant revocation scenarios
- Update policy functions to accept *Registry parameter
MCP tools that accept no parameters return {type: "object"} without a
properties field, which is valid per JSON Schema spec. However, OpenAI
API strictly requires "properties" on object schemas and rejects
requests with HTTP 400: "object schema missing properties".
Add default empty properties map when type is "object" and the MCP
server provides none.
Co-authored-by: Luvu182 <208665161+Luvu182@users.noreply.github.com>
Security: fix cross-agent MCP tool leak by cloning tool registry before MCP registration.
MCP: enforce mcp_ prefix on all tool names, add cache invalidation on server/grant changes,
add grant management endpoints, add group:mcp policy support for per-agent allowlisting.
Skills: persist full YAML frontmatter, auto-promote/demote visibility on grant/revoke,
simplify versioning, handle ZIP wrapper directories, expand tilde in skillsDir path.
Fixes: wrap DeleteSkill cascade in transaction, use atomic NOT EXISTS for revoke-demote,
create cancel context before storing server in map.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Multi-agent AI gateway with WebSocket RPC, HTTP API, and messaging channel integrations.
Go port of OpenClaw with multi-tenant PostgreSQL, per-user isolation, security hardening,
and production observability.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phan Thong IT