| name | self-assess |
| description | Analyze the project against vision.md and spec.md to find gaps, bugs, and improvement opportunities |
| tools | ["bash","read_file","list_files","search"] |
Self-Assessment
You are assessing the project against its vision and specification.
Process
- Read vision.md and spec.md completely
- Read the current project state — list files, read key modules
- Compare spec features vs implementation:
- Which features from spec.md are implemented?
- Which are partially done?
- Which haven't been started?
- Try using the project. Run it, test edge cases:
- Does it start without errors?
- Do the implemented features work as specified?
- Are there crashes, bad error messages, or missing functionality?
- Check JOURNAL.md. Have you tried something before that failed?
What to look for
Spec Fidelity (Gap Analysis)
- Claimed vs. actual — For each [x] item in spec.md, does the implementation
actually deliver what the spec describes? A passing test doesn't mean the feature
is complete. Check: does it work from the user's perspective?
- Shallow implementations — Features that technically exist but lack depth. E.g.,
"error handling" that catches exceptions but shows generic messages instead of the
helpful, specific guidance described in the UX principles.
- Missing connections — Features that work in isolation but aren't wired into the
chat flow as the spec intends. The vision says "chat is the primary interface" — can
every feature be triggered through conversation?
Quality Audit
- Test coverage gaps — Run coverage reports. Which modules are below 90%? Which
have zero coverage? Which have tests that only cover the happy path?
- E2E coverage — Map each spec phase to Playwright tests. Which user journeys
have no E2E test? Which have tests that skip critical steps?
- Integration boundaries — Are there untested boundaries between backend modules?
Between frontend and backend? Between the chat orchestrator and core logic?
- Error resilience — What happens with bad input? Empty CSV? Single row?
All-null columns? Non-UTF8 encoding? Test these and document failures.
Existing Issues
- Broken functionality — things that used to work but don't
- Missing error handling — silent failures, unhelpful messages
- Missing edge cases — empty input, invalid data, boundary conditions
- UX gaps — confusing behavior, unclear output
Vision Alignment (Innovation Opportunities)
- Re-read vision.md — What aspects of the vision aren't yet reflected in the code?
Not just features, but qualities: "delightful," "feels like a smart colleague,"
"moments of surprise."
- User journey friction — Walk through the full flow mentally. Where would a
business analyst get confused, stuck, or bored? Those are innovation opportunities.
- External integrations — What models, data sources, or tools could make the
platform more useful for the vision's target user?
Output
Write findings as a prioritized list:
SELF-ASSESSMENT Day [N]:
Spec coverage: X of Y features [x] — Z verified working end-to-end
Test coverage: backend X%, frontend X%
E2E scenarios: X of Y critical paths covered
Gap Analysis:
- [SPEC ITEM] — Status: [working | shallow | broken] — [details]
Quality Issues:
1. [CRITICAL/HIGH/MEDIUM/LOW] Description
Innovation Opportunities:
1. [Description] — traces to vision: "[relevant vision quote]"
Recommended Focus This Session:
- Quality: [specific hardening task]
- Innovation: [specific expansion task]
- Balance recommendation: [quality-heavy | innovation-heavy | 50/50] based on recent sessions
Then decide what to tackle this session. Check BACKLOG.md for coordination
with other bot instances before committing to a focus area.
