brain-showcase

Assembles a single audit page - state/docs/brain-showcase-<date>.md - that answers the question "what has the brain actually produced?" Reads existing logs and filesystem only. No LLM calls. No skill re-runs. Empty categories are reported explicitly so the audit is honest.

This skill is the receipts layer for the ## Counts block Robert cares about: images it makes, drafts it writes, meetings it closes, mining it produces, pitfalls it has learned, self-rewrites it has performed.

Steps

1. Parse --date arg (default: today UTC).
2. Scan image artifacts:

• /tmp/snappy-image-<slug>/<slug>/state.json (live runs of the

image skill).

• state/log/skool-partner-cta/*.md (image_url: + image_prompt:

frontmatter from skool CTA posts).

• Recent PNGs on disk (30-day window, known tmp dirs only).

1. Scan draft artifacts:

• state/log/ray-drafts.ndjson (Ray weekly updates).
• state/log/skool-daily-post/*.md.
• state/log/older-client-outreach/*.md.
• state/log/referral-asks/*.md.

1. Latest-per-id collapse on state/log/followups.ndjson (meeting

commitments and their resolution status).

1. Sample the three most-recent state/log/content_atoms-*.ndjson files

as mining output.

1. Walk every state/skills/*AGENTS.md loader and extract the first

bullet under ## Known Pitfalls; sort by mtime and take the top 10.

1. git log --grep "fixer: self-write\|regen\|self-rewrite" --since="14 days ago".
2. Write the showcase doc to state/docs/brain-showcase-<date>.md.
3. Log one row to state/log/chain.ndjson and one eval row.

Trigger

npx tsx state/bin/brain/showcase.ts --date 2026-04-18

Default date is today. The doc is safe to regenerate - it overwrites its own output path.

Honesty rule

If a category has zero artifacts (e.g. no image generations in the last 30 days, no followup rows, no pitfall bullets), the section MUST say so explicitly. Empty categories are not fabricated into filler text. Honest empty beats manufactured signal.

Redaction

Emails and loose name@domain.tld patterns are redacted to <name>@<domain>.tld style before being written into the markdown. Commitment text from followups is also run through the same redactor.

Eval

Actor: state/bin/brain/showcase.ts - reads logs + filesystem and writes markdown.

Auditor: shape gate inside the same script that reads the file back after writing and confirms the ## Counts block is present plus the 🎨 Images section header. Score is written to state/log/evals.ndjson with skill: "brain-showcase".

Actor ≠ auditor constraint: the actor is the code path that walked the filesystem and rendered the markdown; the auditor is the post-write independent re-read of the same file. The audit is NOT the return value of the write call.

const shaped =
  existsSync(outPath) &&
  readFileSync(outPath, "utf8").includes("🎨 Images") &&
  readFileSync(outPath, "utf8").includes("## Counts");
score("brain-showcase", run_id, {
  score: shaped ? 1.0 : 0.0,
  primary_issue: shaped ? null : "shape-gate-failed",
  mode: "auto",
  notes: `counts=${JSON.stringify(counts)}`,
});

Outputs

• state/docs/brain-showcase-<date>.md - the showcase doc.
• state/log/chain.ndjson - one action row.
• state/log/evals.ndjson - one eval row.

Known Pitfalls

• Filesystem-wide find for PNGs is slow and noisy. The script only walks

known tmp dirs (/tmp/snappy-image-*, /tmp/bws-images, /tmp/ray-images, /tmp/image-redesign, /tmp/image-spike). Other locations are invisible to this audit on purpose - add them explicitly if a new image pipeline starts writing elsewhere.

• The pitfall walker takes only the first bullet per agents.md. A skill

with many pitfalls still shows up as one line. That is the recency trade-off; adjust findRecentPitfalls(limit) if a deeper cut is needed.

• state/log/followups.ndjson is append-only with latest-wins semantics.

Do NOT sort by ts alone; group by id first, then take max ts_resolved or ts_created.

Rubric

criteria:
  - name: doc_present_and_well_formed
    kind: deterministic
    check: "The command 'npx tsx state/bin/brain/showcase.ts' produces 'state/docs/brain-showcase-<date>.md', where <date> is today's date, and this file contains '## Counts' and '🎨 Images'."
  - name: honesty_rule_adherence
    kind: judge
    check: "The generated markdown document explicitly states 'No artifacts found.' or similar for any category that has zero artifacts, rather than omitting the section or fabricating filler text."
  - name: redaction_applied_correctly
    kind: judge
    check: "The generated markdown document redacts email addresses and sensitive commitment text as per the 'Redaction' section, replacing `name@domain.tld` with `<name>@<domain>.tld`."
  - name: correct_artifacts_scanned
    kind: judge
    check: "The content of the 'Images', 'Drafts', 'Followups', 'Mining', 'Pitfalls', and 'Self-Rewrites' sections accurately reflects the specified log files, filesystem paths, and git history as described in the 'Steps' section."

brain-showcase - loader

Per-turn rules for the brain-showcase skill. Full reference: state/skills/brain-showcase/SKILL.md. Do not skip these.

Critical Rules

_(no failures recorded yet - this skill has not produced hard-won rules. Read state/skills/brain-showcase/SKILL.md before invoking.)_

Commands

| ui dashboard | state/skills/brain-showcase/resources/ui.openui | |invoke: see state/skills/brain-showcase/SKILL.md Steps section |eval log: state/log/evals.ndjson (skill: "brain-showcase")

Self-Test

An agent reading this should correctly:

1. [ ] Know which lib/bin artifact backs this skill (or that it is prose-only)
2. [ ] Know what to write to state/log/evals.ndjson after invoking
3. [ ] Know the eval mode (auto / shape / manual) from the .md frontmatter

Self-report

If this loader fell short, append a line:

echo "[$(date -u +%FT%TZ)] brain-showcase: <what was missing>" >> state/log/loader-feedback.log

<!-- kernel-ok: historical attribution for the PID-loader pattern. -->

Self-correcting loader (PID feedback)

This is the rule that makes loaders learn. Ported from snappy-kernel, where it demonstrably worked: the snappy-dom-cartographer skill mapped Skool admin UI once, every later ship-agent inherited the full map for free.

Before you finish, do two things:

1. Fix gaps (P - proportional). MANDATORY. If this loader didn't cover your case - if you had to read another file, run an undocumented command, work around a wrong selector, or discover a quirk - you MUST attempt an Edit to this AGENTS.md before you log.

• Default action: edit this file. Surgical. Add the missing line, table

row, or code example where it belongs. No restructuring.

• Only LOGGED is allowed when: the fix needs >10 lines, spans multiple

files, or requires a structural rewrite. In that case the state/regen/drain.sh queue picks it up asynchronously.

• "I didn't have time" / "it's minor" / "the next agent will figure it out"

are NOT valid reasons. Minor gaps compound - that's the whole point of this loop.

• Concrete example of a valid P-fix: discover agent-browser screenshot

takes path positionally not as --output <path>. One line in a table. Edit it in. Do not just log it.

• The goal: the next agent never has to leave the loader.

2. Log the result.

echo "[$(date -u +%FT%TZ)] brain-showcase: <what was missing or fixed> [FIXED|LOGGED]" >> state/log/loader-feedback.log

• FIXED = you patched this loader inline (P-fix).
• LOGGED = too large for inline; the PostToolUse enqueue + Stop-hook drain

will rewrite the loader from scratch on next session-end.

Do not skip this. Every agent run must leave the system better than it found it. The loader is the setpoint; you are the sensor; the gap is the error signal; closing the gap is the correction.

OpenUI Resource

• Skill-owned OpenUI Lang resource: state/skills/brain-showcase/resources/ui.openui. Read it before rendering or editing this skill's generated component surface.
• Treat this resource as a first-class artifact of the skill, not a generic chat response. Improve it when the skill's user-facing output needs to become richer.
• System resources compose OpenUI primitives and inherit SnappyChat tokens. Use ui_contract: branded in SKILL.md only for deliberate platform or client visuals.