Contract redline scan
Skill megandmartin/agent-skills-repo/skills/business-ops/contract-redline-scan
75 production-grade agent skills for Hermes Agent + Paperclip — research, write, organize, earn, and run an AI workforce. Every skill passes a QA gate with hard safety rails. Built by Gen AI Hub.
npx -y skills add megandmartin/agent-skills-repo --skill contract-redline-scanAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 14 days oldThe repository was created 14 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Scans a contract and flags risky clauses — IP assignment, indemnity, auto-renewal, exclusivity, payment terms — into a structured issue list for a human lawyer to review. Use when the user says "review this contract", "check this agreement", "any red flags in this NDA/MSA/SOW", or pastes/attaches a contract. Don't use for drafting the commercial offer itself — use proposal-builder — and never as a substitute for legal advice.
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
5.9 KB, ~1.3k tokens by cl100k_base, as published. Nobody here has run it
Contract Redline Scan
This is issue-spotting, not legal advice. This skill helps a business owner walk into a lawyer conversation prepared — it does not replace the lawyer. Every scan output states this, and no contract gets signed on the strength of this scan alone. A human lawyer signs off.
Reads a contract end to end and produces a prioritized flag list across the five clause families that most often burn small operators: IP assignment, indemnity, auto-renewal, exclusivity, and payment terms. The standard: every flag quotes the exact clause text and page/section, so the lawyer's billable time goes to judgment, not hunting.
When to Use
- User received a contract (NDA, MSA, SOW, partnership, platform ToS) and wants to know what to worry about before the lawyer call.
- User asks "anything scary in here?", "check the payment terms", "does this take my IP?".
- Not for: legal advice, negotiating on the user's behalf, or judging whether to sign — that is the lawyer's job. Not for writing the deal itself — use
proposal-builder.
Quick Reference
| Action | Command / Call |
|---|---|
| Check the text is readable | head -c 500 contract.txt — real sentences, not binary garbage |
| Keyword sweep | `grep -n -i -E "assign |
| Count sections found | grep -c -i "indemnif" contract.txt |
| Save the scan | write redline-scan-<party>-<YYYY-MM-DD>.md next to the contract |
Procedure
- Precheck + rail — confirm you have the contract as readable text (
read_file; if it's a scanned PDF with no text layer, ask the user for a text export and stop). Open your reply with the issue-spotting-not-legal-advice rail verbatim before any analysis. - Full read, then sweep — read the whole document first; grep alone misses clauses written in plain words. Then run the keyword sweep from Quick Reference to catch anything you skimmed past. Note every hit's section number.
- Score the five families — for each, extract the exact clause text and rate 🔴 / 🟡 / 🟢:
- IP assignment — 🔴 if it assigns pre-existing IP, portfolio rights, or "all work product" with no carve-out for your tools/templates.
- Indemnity — 🔴 if one-way against the user, uncapped, or covers the other party's own negligence.
- Auto-renewal — 🔴 if renewal is automatic with a notice window under 30 days or buried terms changes on renewal.
- Exclusivity / non-compete — 🔴 if it blocks the user from serving other clients in their actual niche or survives termination broadly.
- Payment terms — 🔴 if net-60+, pay-when-paid, unilateral setoff rights, or no late-payment remedy.
- Catch-all pass — flag anything else that limits liability asymmetrically, grants unilateral termination, or reads like it survives forever ("perpetual", "irrevocable"). One line each under "Other flags".
- Write lawyer questions — turn every 🔴/🟡 into a specific question the user hands their lawyer ("Section 7.2 assigns 'all deliverables and related materials' — can we carve out pre-existing tools?"). Questions, not conclusions.
- Deliver — fill the template, save the file, and close by repeating: nothing here is legal advice; the next step is a lawyer, not a signature.
Output Template
# Redline Scan — {Contract name / counterparty}
{Date} · ⚠️ Issue-spotting only, NOT legal advice. A qualified lawyer must review before signing.
## Verdict at a glance
🔴 {n} high-attention · 🟡 {n} discuss · 🟢 {n} looks standard
## Flags
### 🔴 {Family} — §{X.X}
> "{exact clause text}"
Why it matters: {one sentence, business terms}.
Ask your lawyer: {specific question}
### 🟡 ...
## Other flags
- §{X} — {one-liner}
## Clean areas
{Families that came back 🟢, one line each — absence of a flag noted explicitly.}
## Next step
Send this scan + the contract to your lawyer. Do not sign before their review.
Pitfalls
- Grep-only scan misses plain-language traps — "Client shall own everything Contractor creates" contains no keyword hit for "assign". Recovery: the full read in step 2 is mandatory; if you skipped it, go back before scoring.
- Skill drifts into advice ("this is fine to sign", "just strike section 7") — that's practicing law. Recovery: rewrite conclusions as lawyer questions; the verdict line counts flags, it never says "safe to sign".
- Quote paraphrased instead of exact — the lawyer can't find the clause and trust collapses. Recovery: every 🔴/🟡 must contain verbatim text in quotes with a section number; re-pull from the source if you summarized.
- Missing text layer silently truncates the contract — you scan 4 pages of a 12-page PDF and report it clean. Recovery: sanity-check length (page count vs. extracted text length); if they don't match, stop and ask for a full text export.
Verification
- Output opens AND closes with the not-legal-advice / lawyer-signs-off rail
- All five clause families scored, each with verbatim quote + section number for 🔴/🟡
- Every 🔴/🟡 has a concrete lawyer question, no sign/don't-sign conclusions anywhere
- Extracted text length is plausible for the document's page count
- Scan saved as a file next to the contract
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.
Gives 0 of the 12 instructions most legal skills give in ~1.3k tokens
Counted across 234 of the 234 authors here whose files we hold, read 2026-08-07
- Use text operators for text fieldsin 11 of 234, across 6 files
- Consult qualified counsel before usein 11 of 234, across 3 files
- Use PatentSearch API for patent searchesin 10 of 234, across 5 files
- Confirm jurisdiction, employment type, and required clausesin 9 of 234, across 2 files
- Choose a document template and tailor role-specific termsin 9 of 234, across 2 files
- Validate compensation, benefits, and compliance requirementsin 9 of 234, across 2 files
- Add signature, confidentiality, and IP assignment terms as neededin 9 of 234, across 2 files
- Open the implementation playbook for detailed templatesin 9 of 234, across 2 files
- Use TSDR for trademark data retrievalin 9 of 234, across 4 files
- Ask for clarification if required inputs are missingin 8 of 234, across 2 files
- Set the USPTO_API_KEY environment variablein 8 of 234, across 3 files
- Use the uspto-opendata-python library for PEDSin 8 of 234, across 3 files
Said here and by no other author read
- read the entire contract before keyword sweeping
- include exact clause text and section number for every flag
- score each of the five clause families
- flag asymmetric liability, unilateral termination, and perpetual terms
- convert every red and yellow flag into a specific lawyer question
- note clause families that are clean
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.