vt-c-research-implement¶
Implement approved research proposals from Obsidian. Reads proposals with [x] Implement checked and executes them.
Plugin: core-standards
Category: Research
Command: /vt-c-research-implement
Research Implementation¶
Automatically implement approved research proposals from Obsidian.
Workflow Position¶
/vt-c-research-ingest → Review in Obsidian → Check [x] Implement → /vt-c-research-implement
▲ THIS COMMAND
Invocation¶
/vt-c-research-implement # Implement all checked proposals
/vt-c-research-implement --dry-run # Preview what would be implemented
/vt-c-research-implement --yes # Skip confirmation prompts
/vt-c-research-implement --file DATE # Only process specific date's proposals
Configuration¶
proposals_folder: "TOOLKIT_ROOT/intake/pending/from-research"
toolkit_root: "TOOLKIT_ROOT/plugins/core-standards"
processed_folder: "TOOLKIT_ROOT/intake/processed"
daily_notes_folder: "~/Documents/Obsidian Vault/Daily notes"
Execution Instructions¶
Step 1: Scan Proposals¶
- List all files in
intake/pending/from-research/ - Read each proposals file (YYYY-MM-DD-proposals.md)
- Parse markdown to find individual proposals
Step 2: Find Approved Proposals¶
For each proposal, check the Action line:
A proposal is actionable if:
- [x] Implement is checked AND
- [x] Completed is NOT checked
Step 3: Parse Implementation Details¶
Extract the YAML block from each approved proposal:
type: new-skill
name: vt-c-research-implement
location: skills/component-name/SKILL.md
source_reference: path/to/source
dependencies:
- modify: skills/other-skill/SKILL.md
Validate required fields:
- type (required): new-skill, modify-skill, new-agent, new-command, documentation
- name (required): Component name
- location (required for new-*): Target file path
- source_reference (optional): Path to source material for context
Step 4: Dry Run Check¶
If --dry-run flag is set:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Research Implementation - DRY RUN
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Found N proposals with [x] Implement:
1. [YYYY-MM-DD] Proposal: [Title]
Type: new-skill
Location: skills/[name]/SKILL.md
Dependencies: 1 file to modify
2. [YYYY-MM-DD] Proposal: [Title]
Type: modify-agent
Location: agents/[category]/[name].md
No changes made (dry run).
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Then exit without making changes.
Step 5: Execute Implementation¶
For each approved proposal, based on type:
new-skill¶
1. Create skill directory: skills/[name]/
2. Read source_reference for context (if provided)
3. Generate SKILL.md based on:
- Finding and Suggestion from proposal
- Patterns from similar existing skills
- Source material content
4. Create SKILL.md with proper frontmatter
5. Create symlink: ln -sf ../plugins/.../skills/[name] ~/.claude/skills/[name]
modify-skill¶
1. Read existing skill at location
2. Read source_reference for context
3. Apply modifications based on Suggestion
4. Update SKILL.md preserving structure
new-agent¶
1. Determine category from location (security, design, research, etc.)
2. Read source_reference for context
3. Generate agent definition based on:
- Finding and Suggestion
- Similar existing agents in category
4. Create agent markdown file
modify-agent¶
1. Read existing agent at location
2. Apply modifications based on Suggestion
3. Update agent file preserving structure
new-command¶
1. Create command file: commands/[name].md
2. Generate command definition with:
- Frontmatter (name, description, allowed-tools)
- Usage examples
- Reference to skill if applicable
documentation¶
1. Identify target documentation file
2. Read existing content
3. Apply additions/modifications
4. Update file preserving structure
Step 6: Confirmation (Unless --yes)¶
Before each implementation (unless --yes flag):
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Implementing: [Proposal Title]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Type: new-skill
Name: [component-name]
Location: skills/[name]/SKILL.md
Will create:
+ skills/[name]/SKILL.md
+ ~/.claude/skills/[name] (symlink)
Will modify:
~ skills/other-skill/SKILL.md (dependency)
Proceed? [y/N]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Step 7: Mark Completed¶
After successful implementation:
-
Update the proposal in the intake file:
-
Move the proposals file to the processed archive:
If only some proposals in the file were implemented, keep the file in pending/ and only update the Action lines for completed proposals. -
Add summary to daily note (optional, if daily note exists):
Step 8: Update Plugin Counts¶
After all implementations: 1. Count new skills/commands/agents added 2. Update plugin.json if counts changed 3. Update README.md counts if changed
Step 9: Display Summary¶
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Research Implementation Complete
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Implemented: N proposals
New Components:
• Skills: X
• Agents: Y
• Commands: Z
Modified:
• Skills: A
• Documentation: B
Plugin counts updated:
• Skills: 49 → 52
• Commands: 39 → 40
Daily note updated with implementation summary.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Implementation Templates¶
New Skill Template¶
---
name: vt-c-research-implement
description: [Description from proposal]
argument-hint: "[optional args]"
---
# [Skill Title]
[Description paragraph based on Finding and Suggestion]
## Invocation
```bash
/[skill-name]
Execution Instructions¶
Step 1: [First Step]¶
[Instructions based on source material and proposal]
Step 2: [Next Step]¶
[Continue with implementation steps]
Notes¶
Generated by /vt-c-research-implement from proposal: - Date: YYYY-MM-DD - Source: [source_reference]
### New Agent Template
```markdown
# [Agent Name]
[Description based on Finding]
## Role
[Agent's specific role and responsibilities]
## Capabilities
- [Capability 1]
- [Capability 2]
## Workflow
1. [Step 1]
2. [Step 2]
## Notes
Generated by /vt-c-research-implement from proposal:
- Date: YYYY-MM-DD
- Source: [source_reference]
New Command Template¶
---
name: vt-c-research-implement
description: [Description from proposal]
---
# /[command-name]
[Description]
## Usage
```bash
/[command-name] # Basic usage
/[command-name] --option # With option
What It Does¶
- [Action 1]
- [Action 2]
Execution¶
Invoke the [skill-name] skill:
→ Load skill from skills/[skill-name]/SKILL.md
Implementation Details:
Skipping this proposal.
⚠ Proposal "[Title]" has invalid type: "[type]" Valid types: new-skill, modify-skill, new-agent, modify-agent, new-command, documentationSkipping this proposal.
ℹ Proposal "[Title]" has no source_reference. Implementation will be based on Finding and Suggestion only.Proceed? [y/N]
## Safety Measures
1. **Backup**: Before modifying any file, create backup at `~/.claude/vt-c-research-implement-backups/YYYY-MM-DD/`
2. **Preview**: Always support `--dry-run` for preview
3. **Confirmation**: Require explicit confirmation unless `--yes`
4. **Rollback info**: Log all changes for potential manual rollback
## Integration with Toolkit Intake
This skill is designed to work with the centralized intake system:
1. `/vt-c-research-ingest` creates proposals in `intake/pending/from-research/`
2. User reviews proposals and checks `[x] Implement`
3. `/vt-c-research-implement` reads and executes approved proposals
4. Completed proposals are moved to `intake/processed/`
**Related skills:**
- `/vt-c-toolkit-review` -- processes project feedback and RCA imports from `intake/pending/from-projects/` and `intake/pending/from-rca/`
- `/vt-c-research-ingest` -- generates the proposals that this skill implements
---
## Review Findings Analysis Mode
When invoked with `--from-reviews`, this skill analyzes cross-project review data to propose agent prompt improvements.
### Invocation
```bash
/vt-c-research-implement --from-reviews # Analyze and generate proposals
/vt-c-research-implement --from-reviews --dry-run # Preview proposals only
Execution (Review Analysis Mode)¶
Step R1: Load Cross-Project Data¶
- Read
plugins/core-standards/docs/cross-project-review-patterns.md - Scan for per-project
docs/review-history.mdfiles in known project paths - Merge data into a unified view of recurring findings
Step R2: Identify Agent Improvement Opportunities¶
For each pattern with frequency >= 3:
- Identify which agent should catch this pattern type
- Read the agent's current prompt (from
plugins/core-standards/agents/review/[agent].md) - Check if the pattern is already covered in the agent's instructions
- If NOT covered: generate a proposal to update the agent prompt
Step R3: Generate Proposals¶
Create proposals in the standard intake format:
### Proposal: Update [agent-name] to detect [pattern]
**Finding**: Review retrospectives show [pattern] occurring [N] times across [M] projects, but [agent] does not currently check for it.
**Suggestion**: Add a check for [pattern] to the agent's review instructions.
**Implementation Details**:
```yaml
type: modify-agent
name: vt-c-research-implement
location: agents/review/[agent-name].md
source_reference: plugins/core-standards/docs/cross-project-review-patterns.md
Action: [ ] Review [x] Implement [ ] Skip [ ] Completed ```
Write proposals to intake/pending/from-reviews/YYYY-MM-DD-review-analysis.md.
Step R4: Update Tracker¶
After generating proposals, update cross-project-review-patterns.md:
- Mark patterns as "Proposal generated" in the Toolkit Action column
- Log the proposal date and file in the Improvement Log table