MIMIR

Tool annotations

No tools have read-only/destructive annotations

Add readOnlyHint or destructiveHint annotations to every tool so hosts can warn users before invoking.

All four hints declared on every tool

88/88 tools missing one or more hints — memory_add (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); memory_search (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); memory_list_all (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint), +85 more. OpenAI's directory rejects tools where any of the four hints are missing or non-boolean.

For every tool, set all four hints (readOnlyHint, destructiveHint, idempotentHint, openWorldHint) to explicit true/false values that match the handler’s actual behaviour.

Destructive tools are labelled

9 tools perform destructive updates without destructiveHint — memory_add deletes at line 355 (os.remove(os.path.join(MEMORY_DIR, f"{stale['name']}.md"))); memory_delete deletes at line 490 (os.remove(os.path.join(MEMORY_DIR, f"{name}.md"))); memory_clear deletes at line 512 (os.remove(os.path.join(MEMORY_DIR, f"{e['name']}.md")))

Add destructiveHint:true to any tool whose handler calls .delete(), .upsert(), .update(), unlink, rm, DELETE, DROP, REPLACE INTO, or any operation that overwrites existing data.

Shell command execution

6 child_process/subprocess calls in production code — runs shell commands (mimir/vscode-extension/src/extension.ts:592, mimir/vscode-extension/src/extension.ts:705, mimir/vscode-extension/src/extension.ts:726)

Prefer library functions over shell-outs. If you must shell out, ensure all inputs are properly escaped.

Secrets stay with their owner

1 secret/sensitive value flow into network calls (VLLM_API_KEY → dynamic)

Audit where credentials are sent. A NOTION_TOKEN should only reach api.notion.com — never a third-party host.