feat(plugins): add bm-decide and bm-orient skills to Claude Code plugin

[phernandez]

Why

#1124/#1125 brought the Claude Code plugin up to the Codex plugin's coding-session design (shared bm-writing standard, bm-checkpoint, coding_session schema, sessionProfile/repository config, core hook support). Two deliberate gestures remained Codex-only: decision capture and mid-session orientation. This closes that gap so the two plugins have skill parity.

What Changed

• skills/bm-decide/ — deliberate decision capture: rationale, alternatives, consequences, affected work, as a type: decision note written to the bm-writing standard. The deliberate version of the output style's inline reflex — works even when that style is off (previously the only path). bm-remember and the output style now point to it.
• skills/bm-orient/ — deliberate mid-session orientation: active tasks, open decisions, recent session checkpoints, plus coding_session recall for coding profiles — always scoped to the configured repository; an unscoped coding-session query is forbidden, matching the Codex contract. Presents an evidence-backed summary with permalinks. The mid-session counterpart to the SessionStart brief.
• scripts/validate_claude_plugin.py — requires both skills and pins their contracts (Apply the \bm-writing` skill` for bm-decide; the scoped-query rules for bm-orient), mirroring the Codex validator.
• README (bullets + command rows) and plugin CHANGELOG entries.

Verification

• python3 scripts/validate_claude_plugin.py ✓ (now requires 8 skills)
• python3 scripts/validate_codex_plugin.py ✓
• claude plugin validate plugins/claude-code --strict ✓
• uv run pytest tests/test_codex_plugin_package.py tests/cli/test_coding_session_context.py — 23 passed

🤖 Generated with Claude Code

https://claude.ai/code/session_01A2bgrGfWiL4izcjj66u9R8

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 212b3b3aec

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Restrict bm-decide to explicit decision-capture requests

When the Basic Memory output style is enabled, a user saying a normal decision phrase like “let’s go with X” already triggers the output style’s inline decision capture, while the newly added output-style guidance only delegates to this skill when the user explicitly asks to record a decision. This broad skill description tells Claude Code to invoke bm-decide for the same ordinary decision phrases, so those projects can end up with duplicate decision notes for one choice; narrow the trigger to explicit record/invoke requests or otherwise make it mutually exclusive with the output-style reflex.

Useful? React with 👍 / 👎.