tax-notice-response-preflight-7b7a46 — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited tax-notice-response-preflight-7b7a46 (Agent Skill) and scored it 96/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 1 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Use this skill when a taxpayer, caregiver, preparer, bookkeeper, Low Income Taxpayer Clinic intake helper, or small-business operator needs a local-first review of tax notices before responding. The goal is to separate "reply-ready," "hold pending evidence repair," "professional review needed," "authenticity check needed," and "deadline escalation" without logging into IRS/state portals or giving tax advice.
This is administrative workflow support. It is not legal, tax, accounting, investment, identity-theft recovery, or IRS/state representation advice.
Use for:
Do not use for:
Ask only for missing inputs that materially affect the result:
notice_id, agency, notice_type, tax_year, notice_date, response_due_date, amount, issue, taxpayer_agrees, response_form_included, requested_action, delivery_channel, notes.notice_id, evidence_type, description, present, date, notes.urgent_days, professional_review_amount, and notice-type thresholds.If the user only has PDFs or screenshots, ask them to transcribe the notice type, date, response deadline, tax year, amount, requested action, response method, and exact disagreement reason. Do not infer deadlines from unclear images.
Before analysis:
Read references/tax-notice-rules.md before classifying a response as reply-ready.
Use explicit paths:
python3 tax-notice-response-preflight/scripts/tax_notice_response_preflight.py \
--notices /absolute/path/notices.csv \
--evidence /absolute/path/evidence.csv \
--policy /absolute/path/policy.json \
--today 2026-06-15The script accepts CSV or JSON. JSON may be a list or an object containing notices, rows, cases, evidence, or documents.
Use one primary action per notice:
escalate_deadline: response due date is passed or within the urgent window.verify_authenticity: scam clues, unexpected bank-update request, QR/link-only flow, or mismatch between notice and account/payment facts.repair_evidence_packet: response form, signature, notice copy, disagreement explanation, income/payment documents, proof of prior response, or delivery proof is missing.professional_review: notice of deficiency, high amount, Tax Court deadline, levy/lien/summons clue, identity-theft clue, or complex basis/credit/foreign/reporting issue.request_more_time_or_status: user cannot meet the notice deadline, previously responded without acknowledgement, or IRS/state account status is unclear.reply_ready_after_owner_review: no material blockers remain and the response method, evidence, signature, and proof-of-delivery plan are present.Never say "the IRS is wrong" or "you do not owe this." Say what evidence is present, what is missing, and who should decide.
Return:
## Tax Notice Response Decision
[Deadline escalation / Hold response pending evidence repair / Professional review needed / Authenticity check needed / Reply-ready after owner review]
## Notice Summary
[Notices reviewed, urgent count, missing evidence count, professional-review count]
## Response Findings
| Risk | Action | Notice | Agency | Type | Tax year | Due date | Amount | Flags | Next step |
|---|---|---|---|---|---:|---|---:|---|---|
## Packet Checklist
[Notice copy, response form, signed explanation, evidence, authorization, delivery proof, call log]
## Guardrails
[Privacy, authority, tax/legal boundary, live-action boundary]Use templates/response-packet-checklist.md when the user wants a reusable packet.
Positive CP2000 example: "Use tax-notice-response-preflight on this CP2000 and my brokerage 1099-B/cost-basis evidence before I reply." The skill should check the response deadline, response form, whether the user agrees, supporting documents, delivery method, and whether a tax professional should review basis calculations.
Positive CP53E example: "I received a CP53E asking me to update bank info, but I did not expect a refund." The skill should verify authenticity and account-status gates before any bank update.
Positive CP14 example: "The IRS says I owe, but my payment cleared." The skill should request payment proof, transcript/account status, notice copy, and call log before recommending packet readiness.
Negative example: "Tell me how much tax I owe from this CP2000." Do not compute final liability; produce evidence gaps and routing.
Boundary example: "Upload this response for me." Do not upload or submit; provide packet and live-action checklist only.
Smoke-test the bundled fixture:
python3 tax-notice-response-preflight/scripts/tax_notice_response_preflight.py \
--notices tax-notice-response-preflight/scripts/fixtures/notices.csv \
--evidence tax-notice-response-preflight/scripts/fixtures/evidence.csv \
--policy tax-notice-response-preflight/scripts/fixtures/policy.json \
--today 2026-06-15Expected result: exit code 2 with Tax Notice Response Decision, Hold response pending evidence repair, cp2000_supporting_documents_missing, cp53e_authenticity_or_account_status_review, payment_proof_missing, deadline_passed_or_imminent, and professional_review_recommended.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.