generated: '2026-08-04'
method: searched
source: https://github.com/CordialExperience/agentic-cookbook
note: Provider-published Agent Skills. Cordial ships an open-source 'agentic cookbook' of self-contained
  Claude Desktop skill bundles (<recipe>.zip, each carrying SKILL.md plus reference/tool-ergonomics.md
  and reference/audience-query-mechanics.md). The SKILL.md of each bundle is saved here verbatim; nothing
  in this directory is authored by API Evangelist. Every skill is grounded in the hosted Cordial MCP server
  (mcp/cordial-mcp.yml), not in the REST API, so its tool calls are MCP tool names rather than OpenAPI
  operationIds.
runtime:
  surface: MCP
  server: https://mcp.cordial.io/mcp
  client: Claude Desktop (skill upload)
  access: read-only
x-evidence:
  fetched: '2026-08-04'
  repo: https://github.com/CordialExperience/agentic-cookbook
  http_status: 200
  bundles: 16
skills:
- file: cordial-ai-model-profiler.md
  name: ai-model-profiler
  description: Explain and profile a single Cordial AI attribute (a crdl_ai_* field). Use when a marketer
    or CSM asks what an AI attribute means, what its distinct values are and how many contacts fall in
    each, why a large share of the file has no score, or what feeds the model. Triggers include "what
    does crdl_ai_… mean", "unique values + counts for this AI attribute", "why do half my contacts have
    no [propensity/affinity] score", "what data feeds this model". Profiles the attribute itself across
    the whole file (or a stated base), reconciling every slice.
  source_bundle: ai-model-profiler.zip
- file: cordial-audience-ai-breakdown.md
  name: audience-ai-breakdown
  description: Profile an audience you already send to by breaking it down across Cordial AI attributes.
    Use when a marketer asks how many of their audience are price sensitive, high purchase-propensity,
    what the send-time or engagement-momentum distribution looks like, or "break down / profile my audience
    by [crdl_ai attribute]". Takes a base audience and reports its distribution across one or more crdl_ai_*
    dimensions, enumerated deterministically so every slice reconciles.
  source_bundle: audience-ai-breakdown.zip
- file: cordial-audience-count-snapshot.md
  name: audience-count-snapshot
  description: Give a marketer/CSM a quick, LIVE "how many contacts are in this audience right now" count
    — for a named saved audience or an ad-hoc filter — so routine operational asks get answered without
    opening Audience Builder. Use when someone asks "how many contacts are in [audience name]," "what's
    the count for [audience/segment] right now," "audience count for today," "how big is the [X] audience,"
    "how many people match [filter/criteria]," "size of this segment," or "current count for [saved audience]."
  source_bundle: audience-count-snapshot.zip
- file: cordial-audience-overlap-diff.md
  name: audience-overlap-diff
  description: Run set operations across two audiences (or an audience and a loaded list) live and reconcile
    them — intersection, A-not-B difference, complement, rule-by-rule diff of why two audiences differ,
    and external-list overlap (e.g. how many on a suppression list are still subscribed). Use when a marketer/CSM
    asks "why does this audience pull more contacts than the client's file," "how many on this suppression
    list are still subscribed/emailable," "what's the overlap between audience A and audience B," "diff
    these two audiences rule by rule," "how many are in both segments / in A but not B," "reconcile the
    count difference between two segments," or "cross-tab my loyalty members by AI attribute." Cordial
    has no native audience-diff or overlap endpoint, so the answer is composed by hand and reconciled.
  source_bundle: audience-overlap-diff.zip
- file: cordial-capability-adoption-audit.md
  name: capability-adoption-audit
  description: Account-level audit of which Cordial AI models and platform capabilities are ENABLED, which
    are actually being USED (contacts scored + dims targeted in live audiences/content), and which are
    still AVAILABLE to turn on — to prep a QBR or feature review and prioritize activations. Use when
    a CSM/marketer asks "what AI models / features do we have enabled?", "which capabilities are we using
    vs not using?", "feature/capability audit for this account", "what's live and what's next to activate
    / turn on?", "which AI models are we paying for but not using?", "go over all the features — which
    are enabled, which haven't we used yet?", or "prep a capability + adoption review for the QBR".
  source_bundle: capability-adoption-audit.zip
- file: cordial-contact-360-trace.md
  name: contact-360-trace
  description: Trace ONE contact end-to-end — confirm the record exists (plus any duplicate), read its
    consent/subscription state, when it opted out, what suppression/send-governance applies, and which
    sends/journeys its identity qualified for — to explain a single-contact anomaly. Use when a marketer/CSM
    asks "what happened to this contact," "why did this person get 4 emails back-to-back / is there a
    frequency cap," "did my suppression actually fire for this customer," "why can't this contact receive
    push even though tokens look valid," "when and how did this number opt out," "find this complaining
    customer and any duplicate record," "is this email still subscribed," or "trace this one contact's
    history."
  source_bundle: contact-360-trace.zip
- file: cordial-data-dependency-trace.md
  name: data-dependency-trace
  description: When a contact attribute looks broken (a field gone to 0, blank, or wrong), trace it backward
    — attribute → the data job that writes it → the supplement/table it reads → the import feed/batch
    — to localize where the dependency broke. Use when a marketer/CSM/Solutions person types "why is this
    attribute suddenly 0/blank/wrong", "trace this attribute back to its data job and feed", "what data
    job/feed writes this field", "which import populates this supplement", or "did the last import/batch
    fail for this table".
  source_bundle: data-dependency-trace.zip
- file: cordial-experiment-report.md
  name: experiment-report
  description: Report the A/B / subject-line experiment results for a Cordial message (or rolled up across
    its sends) — the per-variant breakdown, the real winner on the metric the marketer cares about (open
    / click / conversion / revenue-per-email, bot-adjusted and channel-correct), stated with the right
    confidence and caveats. Use when a marketer asks "who won the A/B test on [message]?", "show me the
    subject-line test results", "which variant performed best", "did version A or B win", "what's the
    winning subject line and the adjusted open rate", or "aggregate the split-test results across all
    sends of this automation."
  source_bundle: experiment-report.zip
- file: cordial-frequency-volume-revenue.md
  name: frequency-volume-revenue
  description: Prove whether send volume and attributed revenue move together or inversely — send volume,
    frequency (sends-per-contact), and revenue over a window, sliced by engagement tier and compared period-over-period
    / YoY (e.g. H1 vs H2) — as a Cordial-branded QBR artifact. Use when a CSM asks "are we over-mailing
    / can we cut frequency without losing revenue," "volume down but revenue up — show me the trend,"
    "send volume vs revenue," "sends per contact this period vs last," "frequency by engagement tier,"
    or wants a mail-frequency-vs-revenue case for a QBR / strategy review.
  source_bundle: frequency-volume-revenue.zip
- file: cordial-list-growth-churn-trend.md
  name: list-growth-churn-trend
  description: Show subscriber/list growth and net churn (gross subscribes minus unsubscribes) by month
    across a multi-year window that exceeds the platform's native 12-month dashboard limit — overall or
    for a named audience. Use when a marketer/CSM asks "show subscriber growth by month for the last 2
    years," "net subscriber gain/loss by month beyond 12 months," "list growth and churn trend," "subscribes
    vs unsubscribes per month," "year-over-year sign-up growth," "net subscribers added/lost for this
    audience," "the dashboard only goes back a year — pull more history," or "monthly emailable list size
    over time."
  source_bundle: list-growth-churn-trend.zip
- file: cordial-orchestration-performance-report.md
  name: orchestration-performance-report
  description: Build an accurate performance report for a Cordial orchestration (automated journey) —
    per-node sends/engagement over a real date window, node-to-node drop-off explained by each step's
    filters/delays, per-run revenue, and any A/B experiment running inside the journey. Use when a marketer
    asks "how is this journey/automation performing," "results by node/step/send for [journey]," "why
    does step 2 send to fewer people than step 1," "which message in the welcome series wins," "what's
    the test running inside this journey," or "journey stats for the last N days." Chains orchestration
    → message nodes → child sends → performance so the numbers tie out.
  source_bundle: orchestration-performance-report.zip
- file: cordial-personalization-discovery.md
  name: personalization-discovery
  description: Explore what's possible to personalize or segment on in a Cordial account. Use when a marketer
    asks what contact attributes exist, what values an attribute actually holds, what data coverage looks
    like, whether an attribute is wired up/used anywhere, or "what can I personalize on / segment by".
    Self-learns the account's real attributes and values before estimating coverage, then checks where
    each attribute is already used in content and data jobs.
  source_bundle: personalization-discovery.zip
- file: cordial-program-performance-report.md
  name: program-performance-report
  description: Build a Cordial-branded, exec-ready report on a whole program's send performance over a
    period — sends, engagement, and revenue broken down by channel, promotional vs transactional, and
    batch vs automated, with WoW/MoM/YoY comparison. Use when a marketer/CSM asks to "analyze our whole
    program's campaign performance and give me a branded report," wants a QBR / exec dashboard for the
    account, asks "how is our email/SMS/push program performing this month vs last," or to "break down
    sends and revenue by channel" or "compare channel performance."
  source_bundle: program-performance-report.zip
- file: cordial-send-health-diagnostic.md
  name: send-health-diagnostic
  description: TRIAGE a send that looks wrong — stuck, slow, under-delivered, or showing an implausible
    delivered rate — by reconciling its INTENDED audience size against sent/delivered/failed counts, classifying
    the symptom (in-progress vs slow-but-moving vs short-fired vs truly stalled vs benign), and either
    resolving it from account data or producing an escalation-ready packet for Cordial support. Use when
    a marketer/CSM asks "why did this send under-deliver," "this message is stuck processing," "is this
    send stuck or just slow," "it only sent to 96 of my audience," "the batch only went to 3k of 7k,"
    "why does SMS show 5% delivered," "my scheduled campaign never went out," or "is the delivery rate
    real or inflated by failures."
  source_bundle: send-health-diagnostic.zip
- file: cordial-tag-coverage-audit.md
  name: tag-coverage-audit
  description: Audit whether messages are tagged consistently enough to slice program/regional reporting
    — what share of the message corpus carries the tags a dashboard slices by (region, promo-type, channel),
    where the untagged gaps are, and which tag names are near-duplicates that fragment a dimension — so
    a CSM/marketer can decide whether to trust an O1/O12 sliced report before building it. Use when someone
    asks "are my messages tagged consistently enough to slice reporting", "what's our tag coverage across
    sends", "which messages are missing region / promo-type tags", "can I trust a regional report or are
    tags too patchy", "audit our message tagging / naming convention", "how many sends are untagged",
    "before I build a sliced report, check tag coverage", or "are our tags clean enough for a segmented
    QBR".
  source_bundle: tag-coverage-audit.zip
- file: cordial-where-used-lookup.md
  name: where-used-lookup
  description: Reverse-dependency lookup — find every place a given sculpt block, HTML include, contact
    attribute, coupon code, or image is referenced across the account (which messages, orchestrations,
    templates, includes) so a marketer can safely edit or retire it without breaking sends. Use when a
    marketer/CSM/Solutions person types "where is this block/include used", "what uses this sculpt block/template",
    "where is this coupon code used", "which messages/orchestrations reference this attribute", "where
    is {$contact.attribute} used", "what drives the footer / which include builds X", "find everywhere
    this content key is referenced", or "what will break if I change/delete this block/include".
  source_bundle: where-used-lookup.zip