Use these notes to review a concept or read at your own pace. The interactive workspace above adds predictions, editable code, and actual run results.
IDEA 1
Attribution preserves who made what
Attribution identifies the creator and the material being used. In this fictional podcast, each asset record supplies a title, creator name, and source label. Keeping those fields exact matters: rewriting a name to sound more familiar can credit the wrong person, even if the resulting sentence looks polished.
The program's job is formatting and validation, not inventing metadata. If a record is missing or a required field is blank, route the asset to review. This is a useful tool workflow for an agent system because the controller can obtain stable records before generating prose, rather than asking a model to recall ownership from memory.
Record to credit line
- Asset ID. Read the selected fixture record.
- Required fields. Preserve title, creator, and source label exactly.
- Format. Join only the supplied fields.
- Review. Keep incomplete material out of the completed credits.
All creators and assets here are invented. Source labels identify fixture records, not external links.
An exact identity
const creator = "Maya Chen";
// Do not rewrite it as "M. Chang" or infer a missing name.Identity fields are evidence, not creative-writing prompts. The program should copy them faithfully from the designated record.
Think it through: The creator field is blank. What should the credit builder do?
IDEA 2
Credit and permission answer different questions
A credit line says who made an asset. Permission determines whether it may be used under the applicable terms and context. Giving credit does not automatically grant permission. This exercise has a simplified allowedForFixture flag authored for each invented asset; it does not interpret real licenses or provide legal advice.
Check the use flag before formatting a finished credit. If the flag is false, mark the asset not_allowed; if the record is absent, mark missing_record. These outcomes help a student choose another fixture asset or ask a reviewer. The program never downloads real media, clones a voice, or publishes the podcast.
Two different checks
- Record exists. Find the asset before making any attribution claim.
- Use flag. Respect the authored exercise permission.
- Credit complete. Only then format the required identity fields.
The fixture permission flag is a teaching simplification, not a real licensing decision.
Credit does not change the flag
{ creator: "Maya Chen", allowedForFixture: false }
// A known creator does not turn false into permission.Attribution and permission are independent requirements. Satisfying one does not satisfy the other.
Think it through: An asset has a complete credit but its fixture use flag is false. Which output fits?
Put it into practice
Format exact supplied metadata while producing specific review reasons for unusable records.
- Run the incomplete-record case and inspect the starter's unfinished credit.
- Insert the required-field guard.
- Run every case and compare the three review reasons.
- Try an invented missing ID in custom input; do not fill it with guessed metadata.
Your next experiment: Add a second asset with a missing sourceLabel. Which credit can be completed, and which needs review?
Preserve exact attribution and check permission separately; missing metadata is a review task, not a writing prompt.
Key terms
- Attribution
- Identifying the creator and source of material.
- Metadata
- Structured information describing an item, such as its title and creator.
- Provenance
- The record of where material or a claim came from.
Sources and scope
Original Stemtiq teaching, reviewed 2026-09-14. The named researchers and organizations do not endorse this course. Classroom cases are authored exercises, not published findings.
Anthropic: Writing effective tools for agents — with agents
11 September 2025
Design distinct tools with clear parameters, relevant returned information, and evaluations of how the agent actually uses them.
A description or schema does not guarantee the right action. A live tool can return different data for the same arguments as its environment changes.
Lewis et al. · NeurIPS: Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
NeurIPS 2020
Combine information retrieval with generation so a model can use external passages while producing an answer.
The paper uses a trained neural retrieval and generation architecture. A keyword lookup illustrates retrieval but neither reproduces that architecture nor proves a retrieved claim true.
AWS · Amazon Bedrock AgentCore: Policy in Amazon Bedrock AgentCore: Control Agent Interactions
Reviewed 13 September 2026 · undated documentation
Evaluate identity and tool inputs at a gateway before allowing a call. Treat policy authoring, review, enforcement, and decision logging as separate operations.
The gateway governs capabilities routed through it. A classroom approval check is not an AgentCore integration or a complete production authorization system.