Backed by: state/lib/xano.ts (Stripe side is an external-dep phantom - manual cross-check until a Stripe lib lands)

reconcile

Three-way reconciliation. Reads:

• Xano invoices table for $month
• Stripe charges via API for $month
• Bank deposit entries from the manual CSV drop in sources/finance/

Matches by amount + date window ±3 days + client memo. Flags unmatched.

Steps

1. Fetch all three sources.
2. Normalize to {amount_cents, date, client, source_ref}.
3. Match pass 1: exact amount + same week.
4. Match pass 2: exact amount + memo overlap.
5. Emit {matched, unmatched_invoices, unmatched_payments, unmatched_deposits}.
6. Write state/log/reconcile/<month>.md.

Eval

score("reconcile", run_id, {
  score:
    shape_ok && all_three_sources_loaded ? 1.0 :
    shape_ok ? 0.5 :
    0.0,
  matched_count: matched.length,
  unmatched_invoices: u_i.length,
  unmatched_payments: u_p.length,
  unmatched_deposits: u_d.length,
  primary_issue:
    !all_three_sources_loaded ? "source-missing" :
    u_i.length + u_p.length + u_d.length > 0 ? "unmatched-present" :
    null,
});

Unmatched ≠ score 0 - unmatched is the signal the skill exists to surface. Score 0 only on shape failure or missing sources.

Gotchas

• Stripe API pagination must complete; partial fetch = false negatives.
• Bank CSV is manual - if missing, score 0.5 with `primary_issue:

"source-missing"`.

• Amount comparison in cents. Never in dollars.

Rubric

criteria:
  - name: output_file_exists
    kind: deterministic
    check: "The file 'state/log/reconcile/<month>.md' must exist after skill execution, where <month> is the input month."
  - name: all_sources_loaded
    kind: deterministic
    check: "The skill evaluation must report 'all_three_sources_loaded' as true, or 'primary_issue' is not 'source-missing'."
  - name: correct_output_shape
    kind: deterministic
    check: "The skill evaluation must report 'shape_ok' as true."
  - name: unmatched_reporting
    kind: judge
    check: "The content of 'state/log/reconcile/<month>.md' must clearly list and itemize any 'unmatched_invoices', 'unmatched_payments', and 'unmatched_deposits' flagged by the skill, providing context like amount, date, and client/memo."
  - name: amount_comparison_correctness
    kind: judge
    check: "The reconciliation log, if reviewing specific matches, must demonstrate that amount comparisons were performed in cents and not dollars, avoiding floating-point issues."

reconcile - loader

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

Critical Rules

• ALWAYS compare amounts in CENTS. Never in dollars. Float comparison on dollars guarantees false negatives.
• ALWAYS complete Stripe API pagination. Partial fetch = false-negative unmatched payments. Do not bail early.
• ALWAYS treat unmatched as the SIGNAL, not failure. unmatched_invoices/payments/deposits > 0 does NOT score 0.0 - that's the entire reason this skill exists.
• Score 0.0 only on shape failure or missing source. Score 0.5 if bank CSV missing (primary_issue: "source-missing"); the manual CSV drop in sources/finance/ is human-owned and may not be present.
• Match window: ±3 days, exact amount + memo overlap. Two-pass: pass 1 exact week, pass 2 memo overlap.

Commands

| ui dashboard | state/skills/reconcile/resources/ui.openui | |invoke (TS): chain state/lib/xano.ts (invoices) + Stripe API + sources/finance/<month>.csv |verify: shape gate (matched/unmatched arrays present), all three sources loaded |eval log: state/log/evals.ndjson (skill: "reconcile") |output: state/log/reconcile/<month>.md

OpenUI Resource

• Skill-owned OpenUI Lang resource: state/skills/reconcile/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.

Known Pitfalls

• Stripe lib is a phantom (per skill .md "external-dep phantom - manual cross-check until a Stripe lib lands"). Until then, manual cross-check is acceptable but score 0.5.
• The requires: [month] payload key is mandatory. Format YYYY-MM.
• "Unmatched" is per-source: an invoice with no payment, a payment with no invoice, a deposit with no payment. Three independent buckets.

Self-Test

An agent reading this should correctly:

1. [ ] Compare amounts in cents, never dollars
2. [ ] Score 1.0 on a clean run with unmatched > 0 (the signal, not failure)
3. [ ] Score 0.5 (not 0.0) when bank CSV is missing

Self-report

If this loader fell short, append a line:

echo "[$(date -u +%FT%TZ)] reconcile: <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)] reconcile: <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.