Add ai-signals skill (RFM, CLTV, NBA, NBP) - #211
Draft
hanzhenliu wants to merge 4 commits into
Draft
Conversation
Customer-facing skill for Treasure AI's AI Signals (PrecisionML) suite, covering all four signals in one skill with progressive disclosure via references/. Owns the AI-Signals-specific material: the ML Batch API call pattern (POST /v1/runs with solution_name + solution_arguments, 408-while-running status polling, regional endpoints), per-signal input and output schemas, complete .dig workflows, parameter tables, output verification and analysis queries, model-quality interpretation, and troubleshooting. Delegates rather than re-explains: .dig syntax, secrets and scaffold to digdag; the http> external-API chassis to llm-workflow; tdx wf operations to tdx-skills:workflow; table exploration to tdx-basic; td_interval to time-filtering; and activation to parent-segment, segment, activation, and connector-config. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Step 3 now ends by posting a plan summary and getting explicit
confirmation, and step 4 depends on it. Previously the only gate was
data-discovery's Step 4, which confirms the source data but says nothing
about what step 4 is about to write - so the user approved the input and
then got a generated project they never reviewed.
The plan block covers the forward half: prep table schema,
solution_arguments, output schema, the .dig files to be created, and the
train/predict cadence. Cadence moves here from step 7 because it decides
the file layout - splitting train from predict also forces a fixed
model_name, since the references' ${session_id} names do not resolve
across sessions, so deciding after generation means pulling the project
and editing both files.
The plan is posted in the conversation rather than written to the project
directory: tdx wf push uploads the whole directory as one revision, so a
plan.md beside the .dig would be deployed to TD and then drift.
data-discovery's Step 4 becomes a hand-off to that block so there is one
confirmation gate rather than two.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
github-actions
Bot
added
the
tdx:core-skills
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Draft / prototype — opening early for feedback on the setup flow and the reference structure.
Adds the
ai-signalsskill covering Treasure AI's PrecisionML suite: RFM, CLTV, NBA, and NBP. Each signal is a Treasure Workflow that POSTs to the ML Batch API and writes scored results to a TD table.Contents
SKILL.md— solution-choice table, prerequisites, regional endpoints, the submit-then-poll API pattern, and a 9-step setup flow from signal choice through activationreferences/rfm.md,cltv.md,nba.md,nbp.md— per-signal input requirements, complete working.digworkflows, parameter tables, output schemas, verification and analysis queries, and troubleshootingreferences/data-discovery.md— candidate-table search, column mapping (including platform patterns for Shopify/Segment/Salesforce), and data-quality checksLatest commit — plan confirmation gate
The last commit adds a confirmation checkpoint at the end of step 3, before project generation. Previously the only gate confirmed the source data; nothing showed the user what step 4 was about to write. The new plan block covers the prep table schema,
solution_arguments, output schema, the.digfiles to be created, and the train/predict cadence.Cadence moved from step 7 to step 3 because it decides the file layout, not just the schedule. Splitting train from predict forces a fixed
model_name— the references'${session_id}names don't resolve across sessions — so deciding after generation means pulling the project and editing both files.Open questions for review
data-discovery.md?ai-signals-skills.zipis committed at the repo root — intentional, or should it be built rather than tracked?🤖 Generated with Claude Code