rdr-accept

RDR Accept Skill

Accepts an RDR after it passes the gate. This is the author/reviewer decision point between gate validation and implementation.

When This Skill Activates

• User says "accept this RDR", "mark as accepted", "approve the RDR"
• User invokes /nx:rdr-accept
• Gate returns PASSED and user confirms acceptance

Input

• RDR ID (required) — e.g., 003

Behavior

1. Verify gate result — read {id}-gate-latest from T2. Block if outcome is not PASSED.
2. Update T2 (process authority) — set status: "accepted", accepted_date: "YYYY-MM-DD".
3. Update file — change frontmatter status: draft to status: accepted, add accepted_date.
4. Update reviewed-by — set to self if empty.
5. Regenerate README — update {rdr_dir}/README.md index.
6. Stage files — git add modified files.
7. Planning handoff (default: yes) — Auto-detect complexity:
   • Scan for any plan section (Implementation Plan, Approach, Steps, etc.)
   • Count step-like subheadings (Phase, Step, Stage, Part, or numbered ###)
   • If 2+ steps → mandatory; otherwise → default yes (opt-out, not opt-in)
   • Ask: "Invoke strategic planner to build execution beads? (y/n) [default: yes]"
   • If no: Continue — no beads created
   • If yes — execute the full chain (3 sequential dispatches + catalog enrichment, orchestrated by this skill):
     1. Write T1 scratch entry tagged rdr-planning-context: mcp__plugin_nx_nexus__scratch(action="put", content="RDR {id}: planning context for {title}. RDR file: {rdr_file}", tags="rdr-planning-context,rdr-{id}"
     2. Dispatch strategic-planner (Agent tool) — create phased plan with beads. Wait for completion.
     3. Dispatch plan-auditor (Agent tool) — audit the plan against codebase. T1 scratch has rdr-planning-context tag. Wait for completion.
     4. Dispatch plan-enricher (Agent tool) — enrich beads with execution context (+ audit findings from T1), write epic bead ID to T2. Wait for completion.
     5. Catalog links (if catalog initialized): Search for related RDRs and create relates links:
        • mcp__plugin_nx_nexus__catalog_search(query="<rdr-title-keywords>", content_type="rdr")
        • For each result that is NOT this RDR: mcp__plugin_nx_nexus__catalog_link(from_tumbler="<this-rdr-title>", to_tumbler="<related-rdr-tumbler>", link_type="relates", created_by="rdr-accept")
        • Skip silently if catalog not initialized or RDR not yet indexed.
     6. Report chain completion to user.
   • Important: Each dispatch is sequential — do NOT dispatch the next agent until the previous one completes. Do NOT rely on agent-to-agent relay for this chain; this skill orchestrates all three dispatches directly.

Agent Invocation

Steps 1-6 execute directly — no agent delegation. Step 7 (planning handoff) optionally dispatches three agents sequentially, orchestrated by this skill (not by agent-to-agent relay):

1. strategic-planner — creates plan and beads
2. plan-auditor — validates plan against codebase
3. plan-enricher — enriches beads with execution context (+ audit findings), writes epic bead ID to T2

Each dispatch waits for the previous agent to complete before proceeding. The accept skill writes a T1 scratch entry tagged rdr-planning-context before the first dispatch.

Important: Step 2 (T2 status: accepted write) must complete before Step 7 (planner dispatch) to satisfy RDR-024 Guardrail 2 (strategic-planner pre-check warns if RDR not accepted).

Success Criteria

• T2 gate result verified as PASSED before accepting
• T2 metadata updated with status=accepted and accepted_date
• File frontmatter updated to match T2
• README index regenerated
• Files staged via git add
• Planning handoff prompt shown with correct default (phase count heuristic)
• T1 scratch entry written with rdr-planning-context tag before planner dispatch
• Strategic-planner dispatched and completed (if user accepts)
• Plan-auditor dispatched and completed (if user accepts)
• Plan-enricher dispatched and completed (if user accepts)

Agent-Specific PRODUCE

Outputs produced directly by this skill (Steps 1-6):

• T2 memory: Updated status record via memory_put tool: project="{repo}_rdr", title="NNN", ttl="permanent", tags="rdr,accepted"
• Filesystem: Updated RDR markdown (frontmatter status: accepted, accepted_date), regenerated {rdr_dir}/README.md
• T1 scratch: mcp__plugin_nx_nexus__scratch(action="put", content="RDR NNN: accepted YYYY-MM-DD" for ephemeral tracking during multi-step acceptance flow
• T1 scratch: rdr-planning-context tag entry via scratch tool (for plan-auditor successor routing)
• Agent dispatch: sequential chain of strategic-planner → plan-auditor → plan-enricher (optional, user-confirmed, orchestrated by this skill)