Map competition materials

Map Competition Materials

Use the installed graphify skill and CLI as the graph engine. Add the competition-specific scope, privacy, evidence, and interpretation rules below.

Non-negotiable privacy gate

1. Never scan a home folder, Downloads folder, attachment cache, drive root, or workspace root recursively.
2. Resolve the exact corpus directory and inventory its files before extraction.
3. Flag application forms, personal identifiers, contact details, unpublished drafts, API keys, credentials, contracts, and private test data.
4. Default to local code-only extraction when the corpus is mixed or its sensitivity is unknown.
5. Before semantic extraction of documents, PDFs, images, audio, or video, state the exact scope, backend, external-transfer behavior, and outputs.
6. Continue with semantic extraction only after explicit user approval for that corpus.
7. Treat graphify-out/ as potentially sensitive. Never publish or commit it without a separate privacy review.

Preflight

1. Run graphify --version; expect the team-pinned version. Do not upgrade silently.
2. Reuse graphify-out/graph.json for questions instead of rebuilding.
3. Inventory and classify files as official requirements, form, team draft, finalist reference, evidence, code, prototype, or review.
4. Separate public references from private team materials before extraction.

Extraction decision

Use Graphify's local code-only path for source code, installation tests, unreviewed private material, and any corpus without semantic-processing approval. State that document relationships remain incomplete.

Use semantic extraction only after the privacy gate passes. Preserve Graphify provenance:

• EXTRACTED: explicitly present;
• INFERRED: derived connection;
• AMBIGUOUS: uncertain mapping.

Never present an inferred edge as an official fact.

Competition analysis queries

Use scoped graphify query, path, and explain calls to determine:

1. which criterion is supported by which claim and evidence;
2. which claims lack sources, analysis, implementation results, or validation plans;
3. where the form, report, prototype, code, and demonstration disagree;
4. which proposed features lack implementation evidence;
5. which successful-reference patterns are reusable structure rather than copyable content;
6. which assumptions connect the problem, intervention, expected result, and KPI;
7. which dates, units, definitions, implementation states, or performance claims conflict;
8. which missing relationships are most likely to cause scoring loss.

Required interpretation

For every important finding, report:

• finding;
• source files, sections, code symbols, or graph nodes;
• extracted, inferred, or ambiguous provenance;
• evaluation or feasibility impact;
• confidence and reason;
• next action.

Distinguish official fact, team analysis, assumption, target, and future plan.

Output order

1. Corpus scope and exclusions
2. Privacy and processing mode
3. Graph summary and major communities
4. Core relationships
5. Claim-to-evidence gaps
6. Contradictions and implementation mismatches
7. Missing validation
8. Priority revisions
9. Limits

Lead with decisions the team can make, not graph terminology.

Failure handling

• If Graphify is missing, report the dependency; do not install or upgrade without authorization.
• If no supported files are found, report the resolved path and file types checked.
• If semantic extraction needs credentials, keep the local graph and ask before configuring them.
• If output unexpectedly shrinks or files are skipped, do not overwrite trusted results.
• If a source cannot be read, mark the analysis incomplete rather than guessing.

Public sharing

Before sharing outputs:

1. exclude original applications, private drafts, downloaded sources, personal data, and test artifacts;
2. scan for local paths, contact information, credentials, and embedded metadata;
3. share pinned installation instructions instead of third-party source code;
4. attribute Graphify and preserve applicable upstream license notices.