Live market signal: Cursor Forum users are asking for a required first tool / pre-tool hook because .cursor/rules can tell an agent to read MEMORY.md, but cannot prove or enforce that the run initialized context before Grep/Shell/Write.
Artifact shipped in abb2287:
Question for Cursor / Claude Code / MCP users:
- Should a preflight receipt be produced by the MCP guard server, by the client/IDE, or by a lightweight CLI before work starts?
- Which tools should be blocked until preflight: only side-effecting tools (
Write, Shell) or also search/retrieval tools?
- Is a non-secret fingerprint of loaded context enough, or do teams need richer provenance such as rule file version, branch, and timestamp?
Hypothesis: if required-first-tool workflows matter, users will ask for a concrete enforcement/receipt boundary rather than another memory-store feature.
Live market signal: Cursor Forum users are asking for a required first tool / pre-tool hook because
.cursor/rulescan tell an agent to readMEMORY.md, but cannot prove or enforce that the run initialized context before Grep/Shell/Write.Artifact shipped in
abb2287:Question for Cursor / Claude Code / MCP users:
Write,Shell) or also search/retrieval tools?Hypothesis: if required-first-tool workflows matter, users will ask for a concrete enforcement/receipt boundary rather than another memory-store feature.