v0.6.0

[0.6.0] - 2026-04-28

Added

• Added a first-phase optional Scapy integration for uds scan and uds trace. UDS results now report protocol_decoder, negative responses now include named NRC metadata, and when the optional Scapy extra is installed the transaction payloads may include summary-level request/response enrichment without changing the canonical CLI envelope.
• Added canarchy j1939 compare for multi-capture comparison of J1939 PGNs, source addresses, DM1 fault changes, and printable TP identification payloads across two or more recorded captures.
• Implemented canarchy re correlate <file> --reference <ref> to correlate candidate bit fields against a reference time series. Accepts JSON (array or named object with name+samples) and JSONL reference formats. Each candidate reports pearson_r, spearman_r, sample_count, and lag_ms (optimal time offset in [-500, +500] ms). Candidates are ranked by |pearson_r| descending. Structured errors are returned for missing --reference (RE_REFERENCE_REQUIRED), invalid or short reference files (INVALID_REFERENCE_FILE), and insufficient capture/reference time-range overlap (INSUFFICIENT_OVERLAP). Also exposed as the re_correlate MCP tool. Closes #52.
• Added fast-scan path to capture-info for large candump files (> 50 MB). The first and last 64 KB are parsed for timestamps and interface/ID samples; frame count is estimated from file size and average line length. The response payload now includes scan_mode: "full" (exact) or scan_mode: "estimated" (large-file approximation). Closes #163.
• Added automatic frame cap for j1939 summary, j1939 dm1, j1939 faults, j1939 inventory, and j1939 compare on captures exceeding 50 MB when no --max-frames or --seconds limit is specified. Analysis is capped at 500,000 frames and a warning is included in the response. Use --max-frames or --seconds to override. Closes #163.
• Added structured failure handling and cache-path validation to the GitHub-backed skills provider workflow so skills fetch and skills cache refresh reject download failures and manifest path traversal attempts without uncaught exceptions.

Fixed

• Fixed the MCP j1939_tp tool to invoke the current j1939 tp sessions --file <file> CLI surface instead of the older bare j1939 tp <file> form, and tightened MCP argv coverage around that mapping.

Documentation

• Marked the first-phase repository-backed skills provider workflow as a current capability in top-level project documentation so the skills umbrella issue reflects the implemented manifest, provider, cache, and agent-reference surface.
• Audited current operator, design, and test documentation for stale command-status and command-syntax references, aligning reverse-engineering helper status and file-backed examples with the implemented CLI surface.
• Defined the phase-1 agent and MCP integration model for CANarchy skills, including the decision to keep skills CLI-discovered and out of the MCP tool/resource/prompt surface while documenting how agents should select, fetch, and reference skill provenance.
• Clarified the agent workflow policy so non-trivial work is expected to happen on dedicated branches and be delivered through pull requests by default rather than direct pushes to main.
• Added dedicated current-state design and test specs for j1939 monitor and config show, and aligned the surrounding J1939 spec language with the current j1939 tp sessions command surface.
• Defined the first version of the CANarchy skill manifest schema, added matching test-spec coverage, and added canonical example manifests so future skills provider and MCP work can build against a stable contract.
• Documented the new multi-capture j1939 compare workflow across the J1939 design/test specs, command reference, and agent-facing command guidance.
• Clarified the release workflow so GitHub release notes use the full matching changelog version section verbatim instead of a summary.