Skip to content

Claude/calm vheatm integration analysis n9qi1a - #50

Merged
Eilodon merged 2 commits into
mainfrom
claude/calm-vheatm-integration-analysis-n9qi1a
Aug 1, 2026
Merged

Claude/calm vheatm integration analysis n9qi1a#50
Eilodon merged 2 commits into
mainfrom
claude/calm-vheatm-integration-analysis-n9qi1a

Conversation

@Eilodon

@Eilodon Eilodon commented Aug 1, 2026

Copy link
Copy Markdown
Owner

No description provided.

claude added 2 commits August 1, 2026 16:20
Verify the CALM/VHEATM integration analysis against both repos at HEAD
(first-hand source reads, file:line evidence) and turn it into a concrete,
phased adoption plan for CALM.

Findings confirmed on real code: MCP/CLI report-validation parity gap
(mcp_server.py vs report_validator.py), release gates off the publish path
(release.yml), CALM FNV-only hashing (pipeline.rs), lexical-not-evidence
grounding (edit.rs), audit log vs authenticated ledger (telemetry.rs),
reserved Unresolved variant (types.rs), 3-way write-path inventory drift
(format_files), plus corrections (bridge-only ConfirmOnly tier, opt-in
elicitation, VHEATM sandbox/provider adapters, legacy corpus in bundle).

Plan: dual-hash SHA-256 evidence digests, multi-dimensional snapshot
contract, durable content-addressed query/edit receipts, append-only
authenticated ledger, positional evidence acknowledgment, replayable
change lifecycle, stable evidence IDs, read-only provider adapter, scoped
single-use signed approval token, dual-authority fail-closed integration
(prepare-authorize-execute-verify), and a release-qualification barrier.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Rcqrd3iSCm3Kj8WBa1HUqH
Merge the deep technical CALM audit (CALM-A01..A20 + P0/P1/P2 program) with
the verified VHEATM-adoption plan into one workstream program (WS-1..WS-14).

Cross-check the audit's Confirmed findings against real code before merging:
atomic_write PID temp path + no dir-fsync + best-effort perms (edit.rs:477),
reindex-fail-returns-success (edit.rs:602), tool-call-count review freshness
(edit.rs:929), non-constant-time bearer compare (http.rs:92), syntax-diff =
ERROR/MISSING node count with no-grammar->allow-write (edit.rs:459),
detection-only scanner (security.rs), advisory-only reviewing_symbol
(common.rs:430), and the existing strong external_proofs resolution-proof
model (schema.rs:107) — reconciling that CALM's resolution provenance is
strong while its action/edit provenance is the real gap.

Each workstream maps an audit need to a borrowable VHEATM pattern
(lifecycle, provenance, approval-token, tool-receipt, sandbox/providers,
validation-receipt/taint, supply-chain-attestation, RG registry, MCP
trust-split) plus CALM-native tracks the audit adds (durable FS semantics,
HTTP hardening, per-path lock/lease, ANN scale, dependency-aware
incremental, multi-repo). Includes phasing, roadmap, milestone gates,
SLO/KPI, governance, anti-goals, and full finding->workstream traceability.
The earlier adopt-from-VHEATM plan is marked superseded for the roadmap and
retained as the file:line verification source.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Rcqrd3iSCm3Kj8WBa1HUqH
@Eilodon
Eilodon merged commit e42a5d0 into main Aug 1, 2026
7 of 8 checks passed
@Eilodon
Eilodon deleted the claude/calm-vheatm-integration-analysis-n9qi1a branch August 1, 2026 16:57
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants