All authors
Claude Skills by alisonjieli-png
github.com/alisonjieli-png141 skills0 installs9 views
- Balance Project Duty CoverageCheck project duty assignments for missing ownership, incompatible roles and workload conflicts. Use before a plan treats named responsibilities as covered.Votes: 0GitHub stars: 2
- Reconcile Return Credit To Original SaleReconcile a proposed return credit to the original sale, delivered quantity and received return. Use when a credit request may exceed eligible units or the original price.Votes: 0GitHub stars: 2
- Reconcile Subcontractor Delivery To ObligationsCompare subcontractor deliverables with versioned obligations and acceptance evidence. Use when a status report says complete but contractual proof is incomplete.Votes: 0GitHub stars: 2
- Screen Payable For Document MismatchScreen an invoice against order, receipt and supplied payment policy. Use before an invoice is marked payable when quantity, price or identity may disagree.Votes: 0GitHub stars: 2
- Trace Resource Request To Funding WindowTest whether a project resource request fits the approved schedule and funding window. Use when capacity appears available but authorization may arrive too late.Votes: 0GitHub stars: 2
- Track Customer Promises Through Case UpdatesTrack every dated customer promise through case events and evidence. Use when a case appears resolved but a promised update or action remains open.Votes: 0GitHub stars: 2
- Audit As Of Data VisibilitySeparate what was known at a past reporting cutoff from what later corrections say about that period. Use for reproducible historical reports.Votes: 0GitHub stars: 2
- Audit Measurement UnitsCheck measurement units and conversions before comparing or aggregating numeric data. Use when sources mix units or a report changes a value's unit.Votes: 0GitHub stars: 2
- Check Rate DenominatorsVerify that a rate's numerator and denominator use a compatible eligible population and observation window. Use before comparing conversions, defects, or service rates.Votes: 0GitHub stars: 2
- Define Time Zone Reporting WindowTranslate a local reporting period into exact event-time boundaries. Use for daily or hourly metrics that cross time zones or daylight saving transitions.Votes: 0GitHub stars: 2
- Rebuild State From Ordered EventsReconstruct an entity's state from its ordered event history. Use when a current-status table disagrees with an append-only event log.Votes: 0GitHub stars: 2
- Reconcile Snapshot ChangesClassify added, removed, and changed records between two complete snapshots. Use when comparing periodic exports or state captures by stable identity.Votes: 0GitHub stars: 2
- Trace Numeric RoundingTrace where rounding changes a computed result. Use when line totals, subtotals, or displayed values disagree by a small amount.Votes: 0GitHub stars: 2
- Validate Aggregate GrainCheck whether a measure can be summed across the requested grouping dimensions. Use when totals may double count entities or stock snapshots.Votes: 0GitHub stars: 2
- Build A Project Decision Provenance MapReconstruct a project's active decisions from supplied records while exposing conflicting, superseded, and ownerless decisions.Votes: 0GitHub stars: 2
- Design A Bounded Product ExperimentSpecify a product experiment with assignment, primary outcome, guardrails, exposure, and stop rules before any users are enrolled.Votes: 0GitHub stars: 2
- Map Observed Customer Journey FrictionTurn consented session observations into a customer journey map that separates observed obstacles from possible causes and fixes.Votes: 0GitHub stars: 2
- Match A Product Claim To Verified CapabilityCheck one proposed product claim against versioned capability and customer-journey evidence before it appears in public material.Votes: 0GitHub stars: 2
- Normalize Vendor Offers For One DecisionCompare supplied vendor offers for one purchasing decision by making units, exclusions, assumptions, and unanswered terms explicit.Votes: 0GitHub stars: 2
- Screen A Study For A Product DecisionAssess whether a supplied study can inform one product decision by checking population, intervention, outcome, limits, and transfer conditions.Votes: 0GitHub stars: 2
- Separate Recurring Support Causes From Repeat ContactsAnalyze a bounded support-ticket sample without counting the same customer issue repeatedly as independent demand.Votes: 0GitHub stars: 2
- Triage Support Requests Against Approved PolicyClassify a customer support request using supplied policy and case evidence, producing a proposed disposition with explicit holds.Votes: 0GitHub stars: 2
- Adjudicate Model Disagreement With External ChecksResolve disagreement among model answers using task evidence and independent checks. Use when several agents return plausible conflicting decisions and a majority vote would hide uncertainty.Votes: 0GitHub stars: 2
- Audit Semantic Loss In Protocol FallbackCheck whether a proposed protocol or adapter fallback loses required meaning. Use before accepting a lower version or alternate representation at a typed system boundary.Votes: 0GitHub stars: 2
- Challenge Retrieval With Hard NegativesDesign near-match and no-match probes for a retrieval catalogue. Use when search may return plausible but wrong skills, tools, or documents instead of abstaining.Votes: 0GitHub stars: 2
- Compare Model Context Pairs On Frozen CasesDesign a paired evaluation of model choice and context package together. Use when a shorter skill or smaller model may help only in combination with the other.Votes: 0GitHub stars: 2
- Explain Engine Selection From Eligible EvidenceAudit why one engine was selected from registered alternatives. Use when a planner must separate capability eligibility, evidence ranking, and a permitted fallback.Votes: 0GitHub stars: 2
- Map Independent Acceptance EvidenceMap completion claims to evidence with separate provenance. Use when several reports or checks may all depend on the same producer, fixture, or flawed source.Votes: 0GitHub stars: 2
- Reconstruct Causality Across Clock Skewed RecordsReconstruct a defensible partial order from clock-skewed logs and traces. Use when a cross-service incident has timestamps that disagree or events that arrive out of order.Votes: 0GitHub stars: 2
- Separate Shared Build Cache From Task StateClassify files and cache keys before parallel agents share build storage. Use when separate worktrees or sandboxes reuse dependencies but must not share task outputs or credentials.Votes: 0GitHub stars: 2
- PayloadFlag numeric values that lie beyond a declared multiple of the median absolute deviation, per column or per group, and report them without deleting or changing any row.Votes: 0GitHub stars: 2
- PayloadReplace category values only through an explicit reviewed mapping table after case and spacing normalization, and list every unmapped value with its count instead of inventing a mapping.Votes: 0GitHub stars: 2
- PayloadConvert a date column to ISO 8601 using only an ordered list of declared formats, and hold every value that no format parses or that two formats read differently, such as 03/04/2025.Votes: 0GitHub stars: 2
- PayloadTurn numeric text into exact decimal values using the declared decimal mark, grouping mark, currency symbols and percent rule, and hold values that mix conventions instead of guessing.Votes: 0GitHub stars: 2
- PayloadTurn only the declared missing-value tokens, such as n/a, null or a lone dash, into empty cells, report counts per column, and leave zero, false and every other real value untouched.Votes: 0GitHub stars: 2
- PayloadCompare each column of train and test files for missing rates, unseen categories and numeric range shifts, and list columns present in only one file, before any feature work starts.Votes: 0GitHub stars: 2
- PayloadAssign every training row to one of K folds so that no group spans two folds and label shares stay close to the whole set, using a fixed seed and printing a self-check.Votes: 0GitHub stars: 2
- PayloadCreate forward-moving train and validation windows from a timestamp column with a declared gap, so no validation row is older than any training row it is scored against.Votes: 0GitHub stars: 2
- PayloadRead an experiment ledger of per-fold scores and rank runs by mean and spread, marking a gain as not established when it is smaller than the fold-to-fold variation.Votes: 0GitHub stars: 2
- PayloadCompute smoothed target means for categorical columns using only the other folds for each training row and the full training set for test rows, so no row's encoding sees its own label.Votes: 0GitHub stars: 2
- PayloadScan a saved harness session log for assistant turns that describe a tool call in plain text without making a structured call, the failure that stalls small local models, and report the rate.Votes: 0GitHub stars: 2
- PayloadParse pytest, unittest or JUnit XML output into a short list of failing tests with the assertion message and the first project frame, leaving out passing noise and long logs.Votes: 0GitHub stars: 2
- PayloadTurn a ticket exported as Markdown, plain text or JSON into a numbered checklist of acceptance criteria, reproduction steps and named files, and flag a ticket that states no checkable outcome.Votes: 0GitHub stars: 2
- PayloadProduce a bounded tree of a repository that marks source, test, configuration and generated folders with file counts, so a small model can choose where to look without opening every file.Votes: 0GitHub stars: 2
- PayloadList the test files that import, directly or through local modules, the Python files changed in a diff, so a step runs the narrowest relevant tests before the full suite.Votes: 0GitHub stars: 2
- Pre Submission GateDecide, without uploading anything, whether one competition submission file may go to the host for upload: it must be the exact file recorded in an experiment note that used the brief's metric, folds and data version, before the deadline and under the daily limit.Votes: 0GitHub stars: 2
- Read Step AssignmentRead the assignment of the current focused step from its packet in .baltor/step/: task fields, packet files, required result fields and earlier results. Use it first in a fresh session that holds a step packet, through the step packet tools or the local script.Votes: 0GitHub stars: 2
- Write Step ResultSave the result of the current focused step as a new numbered JSON file that is never overwritten, after a check of the required fields and top-level types in the packet's output schema. Use it when the result is ready, or later to publish a better result, through the step packet tools or the local script.Votes: 0GitHub stars: 2
- PayloadRefuse a unified diff that touches files outside the step's allowed path patterns, deletes tests, adds binary or very large files, or edits lock and generated files.Votes: 0GitHub stars: 2
- PayloadScan a diff of test files for removed assertions, added skip or expected-failure marks, widened tolerances and deleted test functions, and report each with its location.Votes: 0GitHub stars: 2