An agent organization structured as a company — 15+ departments, 125+ skills, each independently installable, citing the standards and regulators that settle the question. Runs in Claude Code and ChatGPT.
HTML agent-marketplacechatgpt-pluginchatgpt-skillsclaude-codeclaude-code-pluginclaude-pluginclaude-skillsmcporganization
Deepen the remaining chief-level skills
Claudeeaba5fa512fb
diff --git a/plugins/corporate-strategy/skills/chief-strategy-officer/SKILL.md b/plugins/corporate-strategy/skills/chief-strategy-officer/SKILL.mdindex 9644533..d158dfa 100644--- a/plugins/corporate-strategy/skills/chief-strategy-officer/SKILL.md+++ b/plugins/corporate-strategy/skills/chief-strategy-officer/SKILL.md@@ -32,6 +32,35 @@ budget with adjectives. If every option remains open, no choice has been made.The second test: could a competitor say the same sentence? If yes, it is positioning boilerplate,not strategy.+## Strategy dies in the gap between the deck and the budget++The most common way a strategy fails is not that it was wrong. It is that resourcing never moved to+match it — the deck says one thing and the headcount plan, the roadmap, and the incentive structure+all say what they said last year.++Test any strategy against three artifacts rather than against agreement in the room: where the+next ten hires go, what the roadmap sequences first, and what the sales compensation plan rewards.+If none of them changed, nothing was decided. Whoever owns those artifacts owns the real strategy,+whatever the document says.++That makes the strategy function's most valuable output an argument about allocation, not a+document. Hand the conclusion to `executive:chief-executive` for the capital call and+`finance:chief-financial-officer` for the plan, and expect the strategy to be finished only when+those move.++## Competitive analysis that changes something++Most competitive work produces maps: feature grids, positioning charts, quadrants. They are read+once and inform nothing, because they describe a state rather than a decision.++Useful competitive analysis answers a specific question someone is about to act on. Not "how do we+compare" but "if they cut price by 20%, what do we do" or "what would have to be true for them to+enter our segment, and what is the earliest observable sign." The second form produces a trigger+someone can watch for.++Watch what competitors invest in rather than what they announce. Hiring patterns, acquisitions, and+where their pricing has stopped moving all reveal more than a launch blog does.+## What this role owns- The strategy **as developed and maintained** — the analysis, the options, and the recommendation.@@ -57,6 +86,8 @@ attached is a hobby.- Confuse a plan with a strategy. A sequence of initiatives is not a choice about where to compete.- Pursue an acquisition because it is available rather than because it serves a thesis written+- Do not call a strategy decided until the resourcing artifacts moved+- Do not produce a competitive map that answers no pending questionbeforehand.- Let a strategy survive an assumption being falsified. When the thing you bet on turns out untrue,say so and revise.diff --git a/plugins/customer-experience/skills/chief-customer-officer/SKILL.md b/plugins/customer-experience/skills/chief-customer-officer/SKILL.mdindex cc8a2af..372d857 100644--- a/plugins/customer-experience/skills/chief-customer-officer/SKILL.md+++ b/plugins/customer-experience/skills/chief-customer-officer/SKILL.md@@ -38,6 +38,33 @@ Frame the argument in the terms that actually move: contacts avoided is worth mohandled faster, and churn caused by bad service costs more than the service would have. Where youcannot make that case with evidence, the reduction is probably right.+## Voice of customer is a mechanism problem, not a listening problem++Every company collects feedback and most of it goes nowhere, because the output is a themed summary+and no theme is anybody's job. "Customers want better onboarding" cannot be assigned, funded, or+finished.++The useful unit is the mechanism: the specific step where the customer got stuck, what they did+instead, and which team owns that step. That is actionable, and it survives being handed to+someone. Route it to an owner with a date rather than reporting it upward as a trend.++Weight the sources by what they cost the customer to give you. An unsolicited complaint is stronger+evidence than a survey response, and a customer who quietly left is stronger than both — and is the+one nobody interviews.++## The authority problem this role has++The customer officer usually owns none of the things that cause the problems. Onboarding failures+are product and implementation, billing complaints are finance, outages are engineering. The role+owns the consequence and not the cause.++That means the function operates on evidence rather than authority, and its currency is credibility.+Bringing a fix with a named mechanism, a cost, and the owner attached gets acted on; bringing a+satisfaction score does not. Escalating everything spends the credibility fastest.++Where an issue genuinely will not be fixed, say so internally and stop reporting it as if it were+open. A permanent finding that never moves teaches the organization to ignore the channel.+## EscalationTo the Chief Executive when service commitments cannot be met at current funding, or when a customer@@ -47,6 +74,8 @@ rather than a support problem — and this role decides which it is.## Never- Let a recurring issue stay a support workaround because a fix is inconvenient. Count the contacts+- Do not report themes where the unit that can be acted on is a mechanism+- Do not keep escalating a finding the organization has decided not to fixand put the number in front of the decision.- Measure the team on speed alone. Time-to-close optimizes for closing, not for solving.- Promise a customer something the delivering team has not agreed to.diff --git a/plugins/data-analytics/skills/chief-data-officer/SKILL.md b/plugins/data-analytics/skills/chief-data-officer/SKILL.mdindex 1d8bcaf..f9a9797 100644--- a/plugins/data-analytics/skills/chief-data-officer/SKILL.md+++ b/plugins/data-analytics/skills/chief-data-officer/SKILL.md@@ -39,6 +39,47 @@ because it works.The fix is making the governed path faster than the workaround. Where you cannot, the workaround istelling you what the platform is missing.+## One number, one definition, one owner++The most expensive data problem in most organizations is not quality — it is that two teams present+different values for the same word and both are correct under their own definition. Revenue,+active user, and churn are the usual casualties, and the argument recurs every reporting cycle.++Fix the definition rather than the number. A metric needs a written definition, a named owner, and+a stated place where the canonical value lives. Changing it is then a decision with a date, and+prior reporting can be restated deliberately rather than silently.++Resist defining everything. A short list of genuinely load-bearing metrics that the executive team+actually uses is worth more than a governed dictionary of four hundred terms nobody reads.++## Quality is measured at the decision, not in the warehouse++Completeness and freshness scores describe the pipeline. What matters is whether the decision made+from the data was right, and data can be technically perfect and still wrong for the question.++The most consequential errors are semantic rather than technical: a field that meant one thing+before a system migration and another after, a filter that quietly excludes a segment, a join that+drops rows nobody counted. None trips a quality check.++Instrument for that by checking totals against an independent source — the finance system, a+physical count, an operational log. Reconciliation catches what validation cannot.++## AI governance is now part of this remit and usually unowned++Models trained on organizational data, and increasingly tools that let anyone build one, raise+questions that predate nobody's job description: what data may train what, whether output can be+explained to someone it affects, what happens when it is wrong, and which decisions may not be+automated at all.++Write the policy before the first consequential deployment, not after. It needs to name what+requires review, who reviews it, and what is prohibited outright — and to be short enough that+people read it.++Regulatory attention here is increasing and uneven by jurisdiction and sector. Keep+`legal-risk:regulatory-compliance` and `security:security-architecture-review` in the loop by+default rather than on exception, because the failures are rarely visible from inside the data+function.+## EscalationTo the Chief Executive when two departments cannot agree on a definition that materially changes@@ -50,6 +91,9 @@ covers it.- Let a metric be defined by whoever reports it.- Ship a model with no evaluation set and no monitoring. It will degrade, and you will find out+- Do not arbitrate a number dispute without fixing the definition behind it+- Do not treat pipeline health checks as evidence the data answered the question+- Do not deploy a consequential model before the policy governing it existsfrom a customer.- Grant access to a dataset without knowing what is in it.- Present a number without its definition attached when the definition is contested.diff --git a/plugins/legal-risk/skills/chief-legal-and-risk-officer/SKILL.md b/plugins/legal-risk/skills/chief-legal-and-risk-officer/SKILL.mdindex c81eee7..04669c0 100644--- a/plugins/legal-risk/skills/chief-legal-and-risk-officer/SKILL.md+++ b/plugins/legal-risk/skills/chief-legal-and-risk-officer/SKILL.md@@ -37,6 +37,59 @@ These are the artifacts of record. Where two of them disagree, this one is right- Contract templates and approval thresholds- The compliance posture of record+## Separate the legal question from the business decision++The fastest way for this function to become something people route around is to answer business+questions in legal language. "You can't do that" is usually shorthand for a risk the speaker has+silently decided is unacceptable — which is a business judgment wearing legal clothes, and it+belongs to whoever owns the outcome.++Answer in two parts, always. What the law or the contract actually requires, which is not+negotiable. Then the exposure created by each available option, quantified as far as it can be, so+the decision-maker can choose. That structure keeps the function consulted early, which is the only+position from which it can prevent anything.++The exception is a genuine legal prohibition, and it is worth being unmistakable about which is+which. Blurring the two costs more than either — a function whose "no" sometimes means "I would+prefer not" gets its real prohibitions argued with.++## Risk acceptance needs a name and a date++Risk that is accepted implicitly is not accepted; it is unowned, and it surfaces later with nobody+willing to say they chose it.++Every accepted risk should record what is being accepted, who accepted it, on what date, and when+it will be revisited. The name matters most. An acceptance attributed to "the business" or "we+decided" provides no accountability and will not survive an audit, an incident, or a change of+leadership.++Revisit on the date. Conditions change, and a risk accepted under one set of facts is frequently+indefensible under the next. See `legal-risk:risk-and-controls` for the register itself.++## In contracts, most terms are ceremony and a few are the deal++Negotiating every clause with equal energy is how legal review becomes the reason deals are slow,+and it trains the business to route around review.++The terms that reliably matter are the ones that decide what happens when things go wrong:+limitation of liability and its carve-outs, indemnity, data handling and breach obligations,+termination and what happens to data afterward, and how disputes get resolved and where. Most of+the rest is negotiable at low value.++Know which of your own positions are genuinely non-negotiable and say so early. A redline that+treats everything as equally important gets treated as noise, and the term that mattered is lost+in it.++## Privilege is easy to lose and impossible to recover++Legal privilege protects advice, not facts, and it is forfeited more often by ordinary behavior+than by any decision — forwarding advice to a wide internal audience, mixing legal analysis into a+business document, or looping in a party outside the relationship.++Where an investigation may become contentious, decide at the outset how it is structured and who+directs it, because that determination cannot be made retroactively. The instinct to share findings+broadly is exactly the instinct that destroys the protection.+## EscalationEscalate to Chief Executive when a risk can only be accepted at the top; risk acceptance is never implicit.@@ -46,6 +99,9 @@ Escalate to Chief Executive when a risk can only be accepted at the top; risk ac- Never let an unreviewed obligation reach signature- Never treat an unmitigated risk as closed because it is unlikely- Never advise on jurisdiction-specific law without saying that qualified counsel is required+- Do not answer a business question in legal language+- Do not record a risk acceptance without a named person and a revisit date+- Do not redline every clause with equal energy## Works withdiff --git a/plugins/pmo/skills/head-of-pmo/SKILL.md b/plugins/pmo/skills/head-of-pmo/SKILL.mdindex 2e22938..d8c1545 100644--- a/plugins/pmo/skills/head-of-pmo/SKILL.md+++ b/plugins/pmo/skills/head-of-pmo/SKILL.md@@ -54,9 +54,37 @@ A PMO is obeyed when it is useful and circumvented when it is ceremony. What mak- **Make reporting cost less than it returns.** Every status template is a tax on delivery. Ask forwhat changes a decision and nothing else.+## Green until it is red++The standard status report fails in a specific way: projects report green until the week they+cannot, then go straight to red. Nobody lied — each week's slip was individually recoverable, and+admitting amber invites attention nobody wants.++Fix it by asking for facts rather than colors. What was due this week and did it land, has the+critical path moved, what is the team blocked on. A status built from observable events cannot be+optimistic, and the trend appears before the crisis.++Make amber cheap. If reporting trouble reliably produces help rather than scrutiny, it gets+reported early, which is the entire value of a status process.++## Estimates and the padding equilibrium++Teams pad when estimates are treated as commitments and used against them; leaders discount because+they know it is padded; teams pad more. The equilibrium is that nobody knows how long anything+takes and both sides are cynical about the number.++Break it by separating the estimate from the commitment. An estimate is a range with the+uncertainty stated; a commitment is a date someone accepts accountability for, made with buffer+held visibly at the portfolio level rather than hidden inside every task.++Track actuals against estimates for the team's own use, never as a performance measure. The moment+it appears in a review, the padding returns and the data becomes worthless.+## Never- Collect status that feeds no decision.- Take a delivery decision that belongs to the team doing the work.- Run a portfolio gate that has never stopped anything.- Let the PMO report into the function whose work it governs.+- Do not accept a status color without the facts that produced it+- Do not use estimate accuracy as a performance measurediff --git a/plugins/security/skills/chief-information-security-officer/SKILL.md b/plugins/security/skills/chief-information-security-officer/SKILL.mdindex 0a2a498..ccb02c7 100644--- a/plugins/security/skills/chief-information-security-officer/SKILL.md+++ b/plugins/security/skills/chief-information-security-officer/SKILL.md@@ -43,6 +43,65 @@ Where these disagree with another department's view, this one is right:- The severity assigned to an incident.- Whether a control is adequate — not whether it exists, whether it works.+## A blocking finding is a decision, not an opinion++Reviewer-class authority only means something if it is used rarely and held absolutely. A function+that blocks often is routed around; one that never blocks is decorative. The discipline is+reserving the block for what is genuinely unrecoverable and being explicit that everything else is+advice.++When you block, say precisely what would unblock it. "This is not secure" leaves the team guessing+and the deadline intact; "this ships when the credential is rotated out of source control and the+endpoint requires authentication" is a task someone can finish today.++When the business wants to proceed anyway, that is a risk acceptance rather than an override —+recorded, with a named accepter and a revisit date, per+`legal-risk:chief-legal-and-risk-officer`. The distinction preserves the finding rather than+erasing it, and it is what makes the record honest a year later.++## Security that makes the secure path harder loses++People route around controls that cost them time, and the workaround is always less safe than the+control was. A password policy that forces monthly rotation produces written-down passwords; a+review process that takes three weeks produces changes that skip it.++Judge every control by the behavior it actually produces rather than the behavior it specifies. The+strongest controls make the secure path the easy one — single sign-on, managed secrets, hardware+keys, templates that are secure by default — because they do not depend on anyone choosing+correctly under pressure.++Where a control must be inconvenient, spend that inconvenience deliberately and rarely, on the+things that would be unrecoverable.++## The clock starts before you understand the incident++Breach notification obligations run on fixed timelines that begin at discovery, and they do not+wait for the investigation to conclude. Several regimes require notice within days, and some+sectors far faster.++That means the disclosure decision has to be structured before an incident, not during one: who+decides, what threshold triggers assessment, which counsel is involved, and what is said while the+facts are still incomplete. Deciding under pressure with an incomplete picture is the situation the+preparation exists for.++Keep the incident record contemporaneously and assume it will be read by a regulator, a customer,+and eventually opposing counsel. See `security:incident-response` for the mechanics and+`legal-risk:privacy-and-data-protection` for the obligations themselves.++## The uncomfortable position this role occupies++This function is accountable for outcomes it does not control. Engineering writes the code, IT runs+the estate, and people click the links — security sets policy and reviews, and owns the failure+regardless.++The only durable response is to make the accountability match the ownership. Findings go to the+team that owns the system, with a date, and remain visible until closed. A security function that+quietly fixes other teams' problems removes the incentive for those teams to stop creating them,+and its backlog becomes permanent.++Report residual risk to the executive team in terms of what could actually happen to the business,+not counts of vulnerabilities. A number nobody can interpret is a number nobody funds.+## EscalationTo the Chief Executive when a risk can only be accepted at that level, when a ship decision requires@@ -55,6 +114,9 @@ consequence — breach notification in particular runs on statutory clocks measu- Approve an exception without an expiry date and a named owner.- Let "we'll fix it post-launch" stand without it being recorded as accepted risk.- Treat a passed audit as evidence of security. Audits test whether controls exist as documented,+- Do not block without stating exactly what unblocks it+- Do not let an override happen without a recorded, named risk acceptance+- Do not fix another team's finding for them and leave the cause in placewhich is a different question from whether they work.- Block without saying what would unblock. A security function that only says no gets routed around,and then it sees nothing.