skill-implementer

# Implementer Skill Thin wrapper that delegates general implementation to `general-implementation-agent` subagent. **IMPORTANT**: This skill implements the skill-internal postflight pattern. After the subagent returns, this skill handles all postflight operations (status update, artifact linking, git commit) before returning. This eliminates the "continue" prompt issue between skill return and orchestrator. ## Context References Reference (do not load eagerly): - Path: `.claude/context/formats/return-metadata-file.md` - Metadata file schema - Path: `.claude/context/patterns/postflight-control.md` - Marker file protocol - Path: `.claude/context/patterns/subagent-continuation-loop.md` - Continuation loop pattern - Path: `.claude/context/patterns/context-exhaustion-detection.md` - Context exhaustion heuristics - Path: `.claude/context/patterns/file-metadata-exchange.md` - File I/O helpers - Path: `.claude/context/patterns/jq-escaping-workarounds.md` - jq escaping patterns (Issue #1132) - Path: `.claude/context/contracts/phase-closure.md` - depth-first phase closure: close one phase before opening the next (loaded by agent) - Path: `.claude/context/contracts/pre-edit-gate.md` - per-item evidence before applying a mechanical-list edit (loaded by agent) Note: This skill is a thin wrapper with internal postflight. Context is loaded by the delegated agent. ## Trigger Conditions This skill activates when: - Task type is "general", "meta", or "markdown" - /implement command is invoked - Plan exists and task is ready for implementation --- ## Execution Flow ### Stage 1: Input Validation Validate required inputs: - `task_number` - Must be provided and exist in state.json - Task status must allow implementation (planned, implementing, partial) ```bash # Lookup task task_data=$(jq -r --argjson num "$task_number" \ '.active_projects[] | select(.project_number == $num)' \ specs/state.json) # Validate exists if [ -z "$task_data" ]; then return error "Task $task_number not found" fi # Extract fields task_type=$(echo "$task_data" | jq -r '.task_type // "general"') status=$(echo "$task_data" | jq -r '.status') project_name=$(echo "$task_data" | jq -r '.project_name') description=$(echo "$task_data" | jq -r '.description // ""') # Validate status (only block terminal states) if [ "$status" = "completed" ] || [ "$status" = "abandoned" ] || [ "$status" = "expanded" ]; then return error "Task is in terminal state [$status]" fi ``` --- ### Stage 2 + Stage 3: Preflight Status Update and Postflight Marker Source `skill-base.sh` once, then follow `@.claude/context/patterns/skill-preflight-flow.md` in full for Stage 2 (preflight status update) and Stage 3 (marker creation): ```bash source .claude/scripts/skill-base.sh padded_num=$(printf "%03d" "$task_number") skill_name="skill-implementer" operation="implement" ``` This skill supplies the shared block's preconditions: `task_number`, `padded_num`, `project_name`, `session_id`, `operation`, `skill_name`. Follow the shared block's ordering and failure-semantics rules exactly — do not re-inline the `update-task-status.sh preflight` call or the `.postflight-pending` heredoc here. **Note**: `update-task-status.sh preflight implement` also updates the plan file status to `[IMPLEMENTING]`, in addition to the state.json/TODO.md updates `skill_preflight_update` documents generically — this is existing `update-task-status.sh` behavior for the `implement` operation, unchanged by this conversion. --- ### Stage 3a: Calculate Artifact Number Read `next_artifact_number` from state.json and use (current-1) since summary stays in the same round as research/plan: ```bash # Read next_artifact_number from state.json next_num=$(jq -r --argjson num "$task_number" \ '.active_projects[] | select(.project_number == $num) | .next_artifact_number // 1' \ specs/state.json) # Implement uses (current - 1) to stay in the same round as research/plan # If next_artifact_number is 1 (no research yet), use 1 if [ "$next_num" -le 1 ]; then artifact_number=1 else artifact_number=$((next_num - 1)) fi # Fallback for legacy tasks: count existing summary artifacts if [ "$next_num" = "null" ] || [ -z "$next_num" ]; then padded_num=$(printf "%03d" "$task_number") count=$(ls "specs/${padded_num}_${project_name}/summaries/"*[0-9][0-9]*.md 2>/dev/null | wc -l) artifact_number=$((count + 1)) fi artifact_padded=$(printf "%02d" "$artifact_number") ``` **Note**: Implement does NOT increment `next_artifact_number`. Only research advances the sequence. --- ### Stage 4a: Memory Retrieval (Auto) Retrieve relevant memories from the memory system to inject into the delegation context. **Skip if**: `clean_flag` is true in the delegation context (from `--clean` command flag). ```bash # Check clean_flag if [ "$clean_flag" != "true" ]; then memory_context=$(bash .claude/scripts/memory-retrieve.sh "$description" "$task_type" "" 2>/dev/null) || memory_context="" fi # memory_context will be empty string if: # - clean_flag is true (skipped) # - memory-index.json missing or empty # - no keywords matched any entries # - script exited with error ``` If `memory_context` is non-empty, it will be injected into the Stage 5 prompt alongside the format specification from Stage 4b. If empty, no memory block is injected. **Literature Detection and Injection (Stage 4a, shared block)** Follow `@.claude/context/patterns/lit-stage4a-flow.md` in full to resolve `--lit` and set `lit_context`: call `literature-lit-flag-resolve.sh`, branch on all six directives (`LIT_DISABLED`, `SUBINDEX_PRESENT`, `GLOBAL_MISSING`, `PROMPT_NEEDED`, `AUTONOMOUS_GLOBAL`, `SPARSE_PROMPT_NEEDED`), issue the real four-option `AskUserQuestion` for the two interactive directives (including the "Search online to ingest" option wired to the STABLE-CONTRACT `literature-ingest-online.sh` bridge), apply the two-checkpoint sparse re-prompt after "Use global corpus now", and take the deterministic `[lit:auto]` autonomous fallback when `orchestrator_mode == "true"` (never calling `AskUserQuestion` in that case). This skill supplies the shared block's preconditions: `lit_flag`, `description`, and `orchestrator_mode` (read from the delegation context; default `"false"` when unset). **Note**: `lit_flag` is independent of `clean_flag`. Using `--clean --lit` suppresses memory retrieval but still injects literature briefing. Literature briefing is gated solely on `lit_flag == "true"`. --- ### Stage 4: Prepare Delegation Context Prepare delegation context for the subagent: `session_id` here is the value received in this skill's args (see Stage 2+3's "supplied preconditions" above), passed through verbatim; never regenerated here. The `sess_{timestamp}_ {random}` shape below is illustrative of that value's format only, not an instruction to construct a new one at this stage. ```json { "session_id": "sess_{timestamp}_{random}", "delegation_depth": 1, "delegation_path": ["orchestrator", "implement", "skill-implementer"], "timeout": 7200, "task_context": { "task_number": N, "task_name": "{project_name}", "description": "{description}", "task_type": "{task_type}" }, "artifact_number": "{artifact_number from Stage 3a}", "effort_flag": "{effort_flag from command, null if not set}", "model_flag": "{model_flag from command, null if not set}", "plan_path": "specs/{NNN}_{SLUG}/plans/MM_{short-slug}.md", "roadmap_path": "specs/ROADMAP.md", "metadata_file_path": "specs/{NNN}_{SLUG}/.return-meta.json" } ``` **Note**: The `artifact_number` field tells the agent which sequence number to use for artifact naming (e.g., `01`, `02`). Summary uses the same round number as the research and plan that preceded it. **Model/Effort Flags**: If `model_flag` is set (haiku, sonnet, opus, fable), pass it as the `model` parameter on the Agent tool to override the agent's frontmatter default. If `effort_flag` is set (fast, hard), include it as prompt context for reasoning depth guidance. > **CRITICAL: No Source Reading Before Delegation** -- Between preparing the delegation context (Stage 4) and spawning the sub-agent (Stage 5), the lead skill MUST NOT read, grep, glob, or analyze source files. The plan file and state.json are the only files the lead reads. All codebase exploration (reading source files, grepping for patterns, using MCP tools) is the exclusive responsibility of the sub-agent after it is spawned. --- ### Stage 4b: Read and Inject Format Specification Read the summary format file and prepare it for injection into the subagent prompt. This ensures the subagent always has the full format specification in its context, regardless of whether it reads the file itself. ```bash format_content=$(cat .claude/context/formats/summary-format.md) ``` The format content will be included as a delimited section in the Stage 5 prompt (see below). --- ### Stage 5: Invoke Subagent **Task-lock heartbeat note**: this skill is a thin wrapper that delegates the entire phase loop to `general-implementation-agent` in a single Agent tool call — it has no per-phase-transition point of its own to hook a `task-lock.sh heartbeat` call into. The heartbeat refresh lives inside `general-implementation-agent.md`'s Stage 4D ("Mark Phase Complete"), which fires once per phase as the subagent progresses through the plan. See `.claude/context/patterns/task-lock.md` for the full contract. **CRITICAL**: You MUST use the **Agent** tool to spawn the subagent. **Required Tool Invocation**: ``` Tool: Agent (NOT Skill, NOT Plan) Parameters: - subagent_type: "general-implementation-agent" - prompt: [Include task_context, delegation_context, plan_path, metadata_file_path, AND the format specification from Stage 4b as shown below] - description: "Execute implementation for task {N}" ``` **Format Injection**: Include the format specification from Stage 4b in the prompt as a clearly-delimited section: ``` <artifact-format-specification> ## CRITICAL: Summary Format Requirements You MUST follow this format specification exactly when writing the implementation summary. Non-compliance will be caught by postflight validation. {format_content from Stage 4b} </artifact-format-specification> ``` Place this section AFTER the delegation context JSON and BEFORE any other instructions. **Memory Context Injection**: If `memory_context` from Stage 4a is non-empty, include it in the prompt as a separate block: ``` {memory_context from Stage 4a -- already wrapped in <memory-context> tags} ``` Place the memory context block AFTER the format specification and BEFORE the task-specific instructions. Do NOT inject an empty `<memory-context>` block when no memories were retrieved. **Literature Briefing Injection**: If `lit_context` from Stage 4a is non-empty, include it in the prompt as a separate block: ``` {lit_context from Stage 4a -- already wrapped in <literature-briefing> tags} ``` Place the literature briefing block AFTER the memory context block (if any) and BEFORE the task-specific instructions. Do NOT inject an empty `<literature-briefing>` block when no literature briefing was generated. **DO NOT** use `Skill(general-implementation-agent)` - this will FAIL. The subagent will: - Load implementation context files - Parse plan and find resume point - Execute phases sequentially - Create/modify files as needed - Create implementation summary - Write metadata to `specs/{NNN}_{SLUG}/.return-meta.json` - Return a brief text summary (NOT JSON) --- ### Stage 5a: Validate Subagent Return Format If the subagent's text return parses as valid JSON, log a warning (v1 pattern instead of v2 file-based pattern). Non-blocking -- continue to read metadata file regardless. --- ### Stage 5b: Self-Execution Fallback Follow `@.claude/context/patterns/skill-self-execution-fallback.md` in full. This skill's success status value for that block's write obligation is `"implemented"`. --- ### Stage 5c: Continuation Loop Init Initialize continuation tracking before entering the postflight loop: ```bash continuation_count=0 max_continuations=3 # Create loop-guard file to track count across potential interruptions task_dir="specs/${padded_num}_${project_name}" cat > "${task_dir}/.continuation-loop-guard" << EOF { "session_id": "${session_id}", "continuation_count": 0, "max_continuations": 3, "created": "$(date -u +%Y-%m-%dT%H:%M:%SZ)" } EOF ``` **Note**: The loop guard ensures that even if the skill is interrupted between iterations, the next invocation can read the count and enforce the limit. --- ## Postflight (ALWAYS EXECUTE) The following stages MUST execute after work is complete, whether the work was done by a subagent (Stage 5) or inline (Stage 5b). Do NOT skip these stages for any reason. ### Continuation Loop The postflight stages below run inside a loop. Each iteration processes the return from one subagent execution. If the subagent returns `partial` with a `handoff_path`, a successor subagent is spawned and the loop continues (up to `max_continuations`). ``` while true; do ``` #### Stage 6: Parse Subagent Return (Read Metadata File) Read the metadata file: ```bash metadata_file="specs/${padded_num}_${project_name}/.return-meta.json" if [ -f "$metadata_file" ] && jq empty "$metadata_file" 2>/dev/null; then status=$(jq -r '.status' "$metadata_file")