v2.6.1 — consumer pm:lint hotfix

Hotfix — pm:lint works out of the box in your project again. Patch release; the status enum, the frontmatter contract, and every command name are unchanged.

Fixed

• Version-parity check no longer fails in projects that use Tandem. Previously, npm run pm:lint failed on every run in a project that merely installs the kit, because the check compared your app's package.json version against the kit version and demanded they match — but they are independent by design. The check now only enforces that match inside Tandem's own repo; in your project it simply passes.

Upgrade

Run npm run pm:update to pull the fix. If you applied a local workaround to validate-frontmatter.js, you can drop it after updating.

v2.6.0

Prod-clean installs and stronger planning gates. This release also realigns the public version number with Tandem's internal release line — the jump from 1.1.0 to 2.6.0 reflects that single shared lineage, not 1.4 majors of breaking change. Upgrading is safe: the status enum, the frontmatter contract, and every command name are unchanged.

Fixed

• Clean installs for everyone. A fresh install now carries the Tandem org identity and the current version — no developer email, internal repo name, or broken hook in your project. Seeded standards docs use the published Tandem namespace and the github:DATA-AI-XYZ/Tandem install URL, and the post-tool / stop hooks run without MODULE_NOT_FOUND on every Write / Edit / Stop.

Added

• Planning & verify-gate guards. refine-backlog refuses to promote a story whose premise about another artefact is false; generated testplans assert real signal instead of restating their own prose; and a verify command piped into | tail / | head (which can never fail) is flagged by the validator.
• Version-parity gate now also covers the kit version, so a drifted version can never silently ship or let the health-check mis-report "update available".
• Discoverability — the public manifest now ships discovery keywords, and the README leads with the qualified name ("Tandem — the Claude Code project-management plugin").

Changed

• Name-independent Command Center. The dashboard detects the Tandem plugin by a stable behavioural signal rather than its marketplace name, so the command-flow panel and drawer timeline render regardless of how the plugin was installed.

v1.1.0 — Founder-facing outcomes, generated docs & brand alignment

Additive minor release over v1.0.0. The status enum, frontmatter contract, and existing command names are unchanged — upgrading is safe.

Added

• Founder-facing outcome lines across the whole plan — every artefact carries a plain-English "what you'll be able to do" line, authored automatically (PRD → feature → story → execution strategy) and surfaced on the Command Center.
• /Tandem:document — generate a documentation set (overview, getting-started, architecture, decisions, features) from what Tandem already knows, rendered to HTML.
• /Tandem:curate-toolkit — rank installed AI tools by fit for the project.
• /Tandem:peer-review — on-demand code review (blocker / major / minor findings, each with a fix).
• /Tandem:start-phase + /Tandem:close-phase — open a phase on its own branch; close it with a retrospective and a gated merge.
• Cross-project portability — configurable folder layout, self-wiring pm:install / pm:doctor, cross-platform Node hooks, tiered CLAUDE.md model.

Changed

• Brand-aligned documentation + Command Center (palette + typography); a published docs site under docs/.
• Non-fatal validator warning tier so advisory checks never block.

Live demo: https://data-ai-xyz.github.io/Tandem/