Agent Skills: Plan content below...

status: in-progress phase: 72 wave: 1 depends_on: [] files_modified: ["src/auth/oauth.ts", "src/auth/tokens.ts"] created: 2026-03-02 updated: 2026-03-02

Plan content below...

## Capabilities

### 1. Parse Frontmatter

Extract frontmatter from a markdown file into structured data:

```yaml
# Input file: .planning/phase-72/PLAN-1.md
---
status: planned
phase: 72
plan_number: 1
wave: 1
depends_on: []
files_modified:
  - src/auth/oauth.ts
  - src/auth/tokens.ts
  - src/middleware/auth.ts
task_count: 4
created: 2026-03-02
gap_closure: false
---

Parsed result:

{
  "status": "planned",
  "phase": 72,
  "plan_number": 1,
  "wave": 1,
  "depends_on": [],
  "files_modified": ["src/auth/oauth.ts", "src/auth/tokens.ts", "src/middleware/auth.ts"],
  "task_count": 4,
  "created": "2026-03-02",
  "gap_closure": false
}

2. Extract Specific Fields

Read individual fields without parsing the entire frontmatter:

get_field(.planning/phase-72/PLAN-1.md, "wave") -> 1
get_field(.planning/phase-72/PLAN-1.md, "status") -> "planned"
get_field(.planning/phase-72/PLAN-1.md, "files_modified") -> ["src/auth/oauth.ts", ...]

3. Update Fields

Update individual frontmatter fields without modifying body content:

update_field(.planning/phase-72/PLAN-1.md, "status", "executed")
update_field(.planning/phase-72/PLAN-1.md, "wave", 2)
update_field(.planning/phase-72/PLAN-1.md, "updated", "2026-03-02")

Uses Edit tool to surgically replace only the target field line.

4. Add New Fields

Add fields to existing frontmatter:

add_field(.planning/phase-72/PLAN-1.md, "executed_at", "2026-03-02T14:30:00Z")
add_field(.planning/phase-72/PLAN-1.md, "executor_agent", "gsd-executor")

Inserts new field before the closing --- delimiter.

5. Remove Fields

Remove fields from frontmatter:

remove_field(.planning/phase-72/PLAN-1.md, "gap_closure")

6. Query Across Files

Find documents matching frontmatter criteria:

query(directory: ".planning/phase-72/", field: "wave", value: 1)
-> [".planning/phase-72/PLAN-1.md", ".planning/phase-72/PLAN-2.md"]

query(directory: ".planning/", field: "status", value: "planned", recursive: true)
-> [".planning/phase-72/PLAN-1.md", ".planning/phase-73/PLAN-1.md"]

Uses Grep to search for field patterns, then validates matches.

7. Validate Frontmatter

Validate frontmatter against document type schema:

# PLAN.md required fields
required:
  - status       # planned|executing|executed|verified
  - phase        # integer or decimal
  - plan_number  # integer
  - wave         # integer (execution order group)
  - task_count   # integer

# PLAN.md optional fields
optional:
  - depends_on      # array of plan references
  - files_modified  # array of file paths
  - gap_closure     # boolean
  - created         # date
  - updated         # date
  - executed_at     # datetime

Document type schemas:

• PLAN.md: status, phase, plan_number, wave, task_count
• SUMMARY.md: phase, status, tasks_completed, commits
• RESEARCH.md: phase, status, approach_count, recommended
• STATE.md: last_updated, session_count, current_milestone
• CONTEXT.md: phase, decisions_count, preferences

8. Batch Operations

Update frontmatter across multiple files:

batch_update(
  files: [".planning/phase-72/PLAN-1.md", ".planning/phase-72/PLAN-2.md"],
  field: "status",
  value: "executed"
)

Tool Use Instructions

Parsing Frontmatter

1. Use Read to load the target file
2. Extract content between first --- and second ---
3. Parse YAML content into key-value pairs
4. Return structured object

Updating a Field

1. Use Read to load the file
2. Locate the target field line within frontmatter
3. Use Edit with the exact field line as old_string
4. Replace with new field value line
5. If field is multi-line (arrays), handle properly

Querying Across Files

1. Use Glob to find candidate files in the target directory
2. Use Grep to search for the field pattern (e.g., ^wave: 1$)
3. For each match, verify it's within frontmatter (not body content)
4. Return list of matching file paths

Validating Frontmatter

1. Use Read to load the file
2. Parse frontmatter
3. Check required fields exist for the document type
4. Validate field types (string, integer, array, boolean, date)
5. Return validation result with any errors

Process Integration

• plan-phase.js - Read/write plan frontmatter (wave, depends_on, files_modified)
• execute-phase.js - Update plan status as execution progresses, read wave assignments for parallel execution
• audit-milestone.js - Query all plans by status for completion checks
• research-phase.js - Write research frontmatter (approach_count, recommended)
• discuss-phase.js - Read phase metadata from roadmap sections

Output Format

{
  "operation": "parse|get|update|add|remove|query|validate|batch",
  "status": "success|error",
  "file": ".planning/phase-72/PLAN-1.md",
  "field": "status",
  "previousValue": "planned",
  "newValue": "executed",
  "frontmatter": {},
  "queryResults": [],
  "validation": {
    "valid": true,
    "errors": [],
    "warnings": []
  }
}

Configuration

| Setting | Default | Description | |---------|---------|-------------| | frontmatterDelimiter | --- | YAML frontmatter delimiter | | strictValidation | false | Fail on unknown fields | | autoUpdateTimestamp | true | Auto-update updated field on writes |

Error Handling

| Error | Cause | Resolution | |-------|-------|------------| | No frontmatter found | File missing --- delimiters | Add frontmatter block or skip | | YAML parse error | Malformed YAML in frontmatter | Fix YAML syntax (check indentation, colons, quotes) | | Field not found | Requested field not in frontmatter | Return null, optionally add field | | Type mismatch | Field value does not match expected type | Coerce if possible, error if not | | Edit collision | Field line not unique in file | Include surrounding context for uniqueness |

Constraints

• Frontmatter must be valid YAML between --- delimiters at the top of the file
• Field updates must not modify body content below the frontmatter
• Array fields must use YAML list syntax (either inline [a, b] or block - a\n- b)
• Date fields must use ISO 8601 format
• Boolean fields must use true/false (not yes/no)
• Query operations are read-only; they never modify files
• Batch operations are atomic per file (each file update succeeds or fails independently)