agentleFS
McGuireKevinMcGuireKevin/cell-site-lease-negotiation-toolkit/.claude/skills/negotiation-playbook/SKILL.md

negotiation-playbook

Skill0 starsChanged 59 days agoMITView on GitHub

Build or revise the negotiating position for a telecom lease — who is across the table and which plays they will run, what to ask for, in what order, what each ask is worth, what the landlord will say and the answer to it, and where the walk-away sits. Use after the audit and the comparator have run, when preparing to open a negotiation, when a landlord has pushed back and the position needs revising, or when asked for negotiation strategy, talking points, counterarguments or a fallback position on a lease. Stage 3.5 of the lease negotiation pipeline.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. prosyo-ai/gtm-skill★ 13same name
    2. nandanosql/all_in_one_pm_skills★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository4 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 2 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Never invent a number: Every figure comes from the comparator, through the deal file, and carries worthNote saying which run.

    line 50
  2. Very specific:

    Never invent market evidence: If marketBenchmark is absent, the entitlement argument on rent is unknown, not "probably above market".

    line 56
  3. Very specific:

    Never record a landlord argument you cannot answer: Every entry in anticipated[] carries a disposition — rebut, price, substitute, give or …

    line 78
  4. Distinctive:

    Answer the consent question before you build a position: Read the lease and every amendment and state, in writing and with the provision …

    line 96
  5. Very specific:

    Before the points, before the arithmetic.

    line 147
  6. Distinctive:

    Use what the audit found. "The escalator has been raising it every year" is an argument.

    line 225
  7. Very specific:

    Use tradeFor. Term is the usual currency, and the escalator floor is usually the cheapest thing to give.

    line 276
  8. Very specific:

    When the landlord has pushed back, do not rewrite history.

    line 281

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Negotiation playbook

The audit says what is wrong with the lease. The comparator says what fixing it is worth. This turns both into a position: what to ask for, in what order, and what to do when the landlord says no.

OutputWhat it is
playbook block in <site>.deal.jsonStructured. Travels with the deal into redline and the final report.
<site>-playbook.mdThe brief a negotiator reads. Prose, ordered for a conversation.

Both, every time. The JSON is what later stages consume; the brief is what someone actually takes into the room.

Before anything else

Read these:

  • The deal file. If audit.triage.recommendation is no-action, stop and say so — there is no negotiation to prepare for, and building a playbook implies one.
  • negotiation/positions.json — the standard objections and counters, keyed by clause category.
  • clauses/impacts.json — what trades against what.
  • clauses/clauses.json — tenantBenefit scores, and the clause that would give effect to an ask.

In positions.json specifically, beyond byCategory: dispositions for what counts as an answer at all, counterpartyTypes for the diagnosis, and nonRentCurrency and packages for what to put on the table instead of a number. Read provenance before quoting any figure from it — the bands are evidence of what a landlord has been told, NOT market rates, and they must never reach a deal file.

The red-team overlay may not be there. negotiation/positions.local.json carries the counterparty's doctrine, the thirteen plays with the tell for each, and the counter-sequence. It is gitignored, because the plays are not secret — landlord-side advisers publish them to their own clients — but the carrier's counters are, and a published counter-sequence converts a prepared negotiator into a predictable one. Read it where it exists and treat the two files as one.

Where it is absent, say so in the brief. One line near the top: the position was built without the red-team overlay, so it names no expected plays and no counter-sequence. Do not quietly prepare a thinner position — a negotiator who knows the file is missing can ask for it, and one who does not know it exists walks in believing they have the whole preparation.

Hard rules

Never invent a number. Every figure comes from the comparator, through the deal file, and carries worthNote saying which run. A playbook that works out its own present values will disagree with the model it came from, and two documents in one deal file disagreeing about money is worse than having neither. Where a point cannot be priced — most terms cannot — leave worth blank and say why. A blank is an honest answer; a plausible number is not.

Never invent market evidence. If marketBenchmark is absent, the entitlement argument on rent is unknown, not "probably above market". The market test exists to produce that finding and guessing it invents the thing it was going to establish.

Respect do-not-raise. Where the audit recorded present-better, put the subject in doNotRaise with the reason and leave it out of points entirely. Raising co-location on a lease that permits it with no revenue share invites the landlord to price something they are currently giving away, and the question cannot be un-asked.

Raise everything in the first round. A term introduced in round three reads as bad faith and reopens what was settled. If a point is worth making, it goes in points now with a rank, even if it ranks last.

Say which arguments are borrowed and which were written for this site. anticipated[].source is positions or site. A reviewer needs to know which parts have been used before.

The walk-away is arithmetic. It is baselinePV — the present value of doing nothing. A counter costing more than that is worse than no deal. Do not soften it into a judgement.

Never write a playbook into the repository. It contains the concession sequence and the walk-away number. Write to deals/ or outside the repo, and set confidential: true.

Never record a landlord argument you cannot answer. Every entry in anticipated[] carries a disposition — rebut, price, substitute, give or drop — and if none of the five fits, the objection does not go in the brief. This is not tidiness. An argument written down with nothing beside it reads, to a negotiator under time pressure at the table, as a position that cannot be met: it demoralises without informing, and it is the one part of a playbook that helps the other side if it is ever read by them.

Where an argument genuinely cannot be answered, the ask is what is wrong. That is the honest conclusion almost every time, and it is a finding rather than a gap: drop the point, or re-rank it so the negotiator spends the round on something winnable. Say so explicitly in the brief — "we are not raising X, because their answer to it is better than ours" is useful intelligence. A point silently left in with a blank objection is not.

Conceding is one of the five answers, not the absence of one. Where the landlord is right, say so quickly and visibly — and name what is taken for it. price is a pricing entry, not a concession entry; a price whose counter does not say what the carrier takes in return is unfinished work, and check-positions.ps1 fails on it in the position file for exactly that reason.

Answer the consent question before you build a position. Read the lease and every amendment and state, in writing and with the provision cited, whether the change already falls inside the permitted use and the equipment exhibit. Paying for a right the carrier already holds is the most expensive error available in this discipline, and it is unforced — it happens because the position was built before the document was read. Where the answer is that consent IS required, say so plainly too; a negotiator who claims a right they do not have loses the argument and the credibility together.

Know the compliance position before opening. Recompute the escalation history, reconcile the installed configuration against the operative exhibits, and check that tax, utility and electricity reimbursements are current. A landlord audit arriving two weeks after your letter is priced against the carrier's urgency rather than its merits, and it only works on a negotiator who cannot answer it. Where the audit block already carries a reconciled rent history, cite it; where it does not, say that the position rests on unverified figures.

Never invoke Section 6409(a) against a landlord. It directs state and local governments. It does not touch a private landlord's contractual consent rights, and landlord-side counsel has already warned them that operators may falsely claim regulatory entitlement to free deployment. Raising it confirms their adviser's framing, forfeits credibility across the whole portfolio, and may be a false statement of law to an unrepresented person. Argue the lease — which is the stronger argument anyway wherever the equipment fits inside the exhibit.

Do not put a threat in the playbook that the carrier will not execute. A termination the carrier has no intention of running is worth less than the credibility it costs, and their advisers have specifically told them that carriers threaten termination on sites they intend to keep. If termination is real, say what makes it real. If it is not, leave it out and argue term certainty instead.

Every date in the brief must be a real one. A date is real when it describes something that happens whether the landlord acts or not: a budget close, a crew allocation, an equipment window, a permit expiry. Write what the carrier will do when the date arrives. A hollow date is not a local failure — it teaches this landlord how to read every date given at every other site, and their guidance already tells them carrier deadlines are theatre.

Obligations rise with an unrepresented landlord, they do not fall. Never state or imply the carrier's negotiator is disinterested, correct any misunderstanding about the role, put the economics in writing, and suggest they obtain counsel. A deal extracted from confusion is a deal that gets re-litigated. Where the landlord IS represented in this matter, substantive contact goes through counsel — and in portfolio work an asset manager dealt with for years elsewhere may be represented at this site. Record the answer in counterparty.represented.

Estimates of value are opinion; the other things are facts. A view on what the site is worth, a statement of position, a valuation estimate — opinion. A comparable that does not exist, a misdescription of what the lease says, a misstatement of what an internal approval is, or a deadline described as fixed when it is not — facts, and false ones. The ethical rule and the tactical rule point the same way here, which is worth saying out loud.

Procedure

1. Diagnose who is across the table

Before the points, before the arithmetic. Fill playbook.counterparty, because the same lease against two different owners is two different negotiations and the difference is not the rent.

The asset class tells you what they are optimising. counterpartyTypes in positions.json has twelve, each with what drives the decision and the approach that follows. A hotel owner's real objection is usually construction disruption, answered with windows and escorted access rather than with rent. An industrial owner will sell ground space cheaply — buy the expansion and the generator rights while they are cheap. A self-storage operator bargains hard over small absolute dollars and may be sold mid-negotiation. An aggregator holding the rent stream by perpetual easement cares about churn risk and not at all about your screening package.

The advisory layer tells you how it will be run. Set advisory:

What you will seeWhat it means
unrepresentedEngages on economics immediately, no range, no document requestsYour obligations rise. See the hard rule.
consultantA published range early, with a citation, plus a demand for drawingsUsually paid on the uplift, so process length serves them
counselChecklists and document points rather than ratesFaster on paper, slower on economics; contact goes through counsel
audit-firmA compliance schedule weeks after your letterA claim timed to your need — see the hard rule on knowing your position

Then name the plays to expect. landlordPlays in the overlay has thirteen, with the tell and the counter for each. Put the likely ids in playsExpected — not as a prediction to act on alone, but so that nothing on the call is the first time the negotiator has heard it. Two are worth naming in almost every brief because they are answered by preparation rather than by argument (and where the overlay is absent, these two still are — they are restated here for that reason):

  • decouple-term — they take the payment and decline the extension. Their guidance is explicit that they need not extend in order to be compensated for a modification. Sequence around it: never condition the amendment payment on the extension. Price the extension as its own transaction on its own logic, which is that term certainty is what makes the stream bankable.
  • audit-timed — see the hard rule. The answer is a self-audit, and it has to exist before the first letter, not after the schedule arrives.

And establish the decision chain, courteously and in writing: who recommends, who approves, what they need in hand, when they meet. Deferred authority is frequently genuine with institutional, religious, public and family owners, which makes the play and the truth indistinguishable from outside. Do not try to expose it — you cannot, and trying insults a counterparty who may be telling the truth. Make it a calendar instead of a contest, and supply an approval package rather than an argument, so what goes up their chain is a decision rather than a discussion.

2. Establish what you are working from

Fill basedOn: the comparator window, discount rate, proposal ids, and the dates the comparator and market test were run. This is what makes staleness detectable later — if the proposals or the window have moved since, the playbook describes a deal that is no longer on the table. Say so at the top of the brief rather than quietly regenerating.

Record baselinePV and targetPV from the comparator run. If the deal file carries no comparator output, ask for it rather than estimating.

3. Build the points

One point per subject worth raising. Sources, in order:

  1. audit.termsAnalysis.findings — absent, present-weaker and unclear, excluding do-not-raise. Rank by priority then gap.
  2. The financial case — where marketBenchmark.flag is above, rent and escalator are points whether or not the terms analysis mentions them.
  3. clauses/impacts.json — subjects the audit never flagged that the above drag in. A term ask without a termination ask is the standard error: a longer term without an exit is worse than the shorter one it replaced.

For each point set ask, entitlement, worth + worthNote, ladder, tradeFor, priority, and anticipated from positions.json plus anything site-specific.

Order them for a conversation, not by value. Open on terms rather than rent: a rent ask invites a rent counter and the conversation never leaves the number. Where the audit found only a financial case and clean terms, say that plainly instead of manufacturing term asks — a landlord who spots a padded list stops believing the real items.

4. Make the arguments site-specific

This is the part positions.json cannot do. The standard counter is a starting point; the version that persuades uses this site's facts.

  • A municipal lessor with a public-safety mandate is not a tower company. Read the lease's own language back to them — a landlord-form lease with an emergency-removal right has told you what it cares about.
  • Use what the audit found. "The escalator has been raising it every year" is an argument; "625 in 1998 is 1,270.50 now, computed across 24 escalations" is the same argument that cannot be waved away.
  • Where a finding is low confidence, do not build a headline argument on it. It will not survive the first request for evidence.

5. Build the structure: packages, and the currency that is not rent

Do not open with a number. A single figure can be rejected without a counter, and their guidance tells them to do exactly that — reject a weak offer rather than countering it, so your anchor does no work while their silence still extracts information. Three complete packages priced to the same present value cannot be dismissed the same way: engaging with any of them reveals their priority, and refusing all three without explanation is visibly unserious in front of their own client.

The packages are comparator proposals, not something this skill invents. Build them in the comparator, price them, and record their ids in basedOn.proposalIds. packages in positions.json gives the three shapes — cash now, stream and certainty, operational — and which kind of owner each is for. Match them to counterparty.type.

Price all three to the same present value before presenting them, and be prepared to execute any of them. A menu whose options are not actually equivalent teaches the landlord to combine the best parts of each, and that is a worse position than having opened with one number.

Then build the menu of things to give that are not rent. Most amendments settle on rent because the negotiator had nothing else on the table. nonRentCurrency in positions.json lists thirteen items that cost the carrier little at a site it intends to keep, and each one answers an objection the landlord's own advisers have primed: structural analysis delivered unprompted, construction windows, access protocol, screening, removal and restoration with a timetable, recorded release of the memorandum, lender fee, reporting, utility separation, a defined premises exhibit.

Lead with them. Giving them early spends down the landlord's reservoir of reasons to hold out for rent. Giving them late, one at a time, under pressure, is the counterparty's term-sheet-creep play running on you — and the items arrive in their redline in exactly that order because they have no line-item cost on the day they are conceded.

Two directional rules for the concession order that follows:

  • Trade bonus dollars for escalator points, never the reverse. The escalator is the largest number in most of these deals and the least examined. Refuse uncapped CPI and open market resets outright — both convert a known liability into an open one. If CPI is unavoidable, collar it and define the index and its substitute.
  • Keep a do-not-trade list and say it is one. Affiliate and successor assignment, lender assignment, emergency access, technology substitution, termination rights. These are capital items priced as such, not drafting points to be given away at the redline stage.

6. Decide the concession order before the call

Fill concessionOrder: what is given, in what order, and what each purchase buys. A concession made in the order the landlord pushes is a concession made for nothing.

Use tradeFor. Term is the usual currency, and the escalator floor is usually the cheapest thing to give.

7. On revision, record what actually happened

When the landlord has pushed back, do not rewrite history. Append to pushback:

  • said — their words, as close to verbatim as the record allows
  • response — what was said back
  • outcome — held, conceded, traded, deferred, unresolved
  • movedTo — the revised position
  • newObjection: true where it was not in positions.json

Then update the affected points[].status. Keep anticipated and pushback apart. One is what we expected before the conversation; the other is what happened. Collapsing them destroys the only signal that says whether the position file is any good.

Flag every newObjection in the brief. Those are the ones worth folding back into negotiation/positions.json, and that is how the file improves.

8. Emit both outputs

The brief follows the order of the conversation:

  1. Where this stands — round, what it is based on, and whether that is stale
  2. Who is across the table — type, advisory layer, decision chain, and the plays to expect
  3. The walk-away number, stated once, near the top
  4. The packages, priced to the same present value, and which one fits this counterparty
  5. The asks in order, each with entitlement, worth, the objection and the answer
  6. What we will concede, and what it buys — non-rent items first
  7. What we are not raising, and why
  8. What changed since the last round, on a revision

Before you finish

  • Every figure traceable to a comparator run via worthNote
  • baselinePV recorded — the walk-away is stated, not implied
  • marketBenchmark absent means the rent entitlement says unknown
  • Every present-better finding is in doNotRaise with a reason, and in no point
  • Every point that will be raised is in this round
  • anticipated[].source set on each — borrowed or written here
  • Every anticipated[] entry has a disposition — and any objection that fits none of the five is out of the brief, with the ask dropped or re-ranked and that decision stated
  • Every price disposition names what the carrier takes in return
  • pushback appended to, never rewritten
  • counterparty.type and advisory set, and represented answered rather than assumed
  • The consent question answered in writing, with the provision cited, before any ask was priced
  • The compliance position known — escalation recomputed, configuration reconciled, arrears cleared
  • Packages built as comparator proposals and priced to the same PV, with ids in basedOn.proposalIds
  • Non-rent currency listed and led with, not held back to concede under pressure
  • No Section 6409(a) argument anywhere in the brief
  • No threat the carrier will not execute, and every date describes something that happens anyway
  • confidential: true, written outside the repository

What comes next

A playbook is a position, not an instrument. Before anything is drafted, run the amendment-drafting skill: it assembles and conforms the document stack, traces who actually holds the interests, fixes the term arithmetic, builds the consent list and turns points[] into the issue list a drafter works from. Its brief names the clauses to select in draft.html and carries a rider for the parts of the instrument the clause library cannot produce.

Not built yet

Live recomputation. The brief carries figures copied from a comparator run. A playbook.html viewer would recompute them the way report.html does, and would make staleness impossible rather than merely detectable. Until then basedOn is the guard, and it only works if it is filled in honestly.

Learning from pushback. Objections marked newObjection are collected but nothing folds them back into positions.json automatically. That is deliberate for now — a position file that grows by itself, without a human deciding the counter is any good, is a file nobody can trust.

If the counterparty changes mid-negotiation

Common in telecom, and it does not simply carry over. Where changes[] records a landlord assignment, merger or rename dated during the negotiation, pushback[] up to that date is the position of a party that no longer holds the interest. It is still evidence about the site and the asks, and it is not a commitment the new owner has made or a concession you can hold them to. Say which rounds predate the change rather than letting the history read as one conversation.

A rent-stream acquirer also negotiates differently from a family landlord: the asset is the payment, so term length and escalators price higher for them than a one-time payment does. Note it in basedOn rather than assuming the ladder still holds.

Data entry for any of this is deal-intake, not this skill.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.