progress-update

Progress Update Skill

Record work completed during a session. Appends to a daily progress file via the devplans.ts script.

When to Use

• End of a productive session
• Before switching to different work
• After completing a milestone
• After subtask-heavy sessions (triage learnings and QUIRKS candidates)
• Multiple entries per day are normal — each gets a timestamped heading

Critical Rule: Never Read the Progress File to Append

The devplans.ts script handles all file mechanics: creation, insertion ordering, section management. Do NOT read the progress file. Instead:

1. Run devplans.ts progress to get metadata and current Next Steps
2. Compose your entry
3. Pipe JSON to devplans.ts progress append

This eliminates context bloat from reading growing files, prevents non-chronological insertions, and ensures the Summary and Next Steps sections stay consistent.

File Location and Naming

Progress files live in {dev-plans}/{project}/progress-updates/YYYY-MM-DD-progress.md.

The script handles file creation. If it's before 6:00 AM local time, the script uses the previous day's date (late-night sessions are continuations, not new days). The output includes dateShifted: true when this happens.

Workflow

Step 1: Get Current State

bun ~/.config/opencode/scripts/devplans.ts progress

Returns:

{
  "file": "~/projects/dev-plans/.../2026-02-16-progress.md",
  "tempFile": "~/projects/dev-plans/.../progress-updates/.tmp-entry-1739745000000.json",
  "time": "14:30",
  "status": "existing",
  "effectiveDate": "2026-02-16",
  "dateShifted": false,
  "nextSteps": ["Commit Phase 4.5 code", "Run data quality audit"]
}

Use nextSteps to understand what was planned — update or replace in your entry. Use tempFile as the write destination in Step 3 (auto-approved path, no permission prompts).

Step 2: Compose Your Entry

Build a JSON object:

{
  "slug": "4.5.3 Execution Simulator",
  "content": "Core backtesting engine — replays historical stock signals as hypothetical options trades.\nThis is where the GO/NO-GO answer comes from.\n\n`runOptionsReplay(signals, params)` iterates signals, determines option type, finds strike\nmatching target delta, prices via synthetic pricer, applies slippage, runs day-by-day exit\nrule simulation.\n\nDiffLoupe: low risk, 3 low findings (all acceptable for internal tooling).\n\n56 tests, typecheck clean, full suite 1906 pass.\n\n### Files Changed\n\n- `src/backtest/options/signal-replay.ts` — replay engine, strike selection, matrix runner (310 lines)\n- `src/backtest/options/signal-replay.test.ts` — 56 tests (397 lines)",
  "nextSteps": [
    "Run remaining Track A experiments",
    "Commit all Phase 4.5 code",
    "4.5.8 (17yih5wg): Data quality audit"
  ]
}