Users and Jobs to Be Done

Users and Jobs to Be Done

Primary personas

Application engineer

Needs to understand and run an unfamiliar repository without manually discovering every runtime, database, port, or environment variable.

Job: When I clone or inherit a project, help me understand what it needs and produce a trustworthy local setup so I can contribute without days of trial and error.

DevOps or platform engineer

Owns delivery and environment consistency across several projects with limited time and incomplete documentation.

Job: Before a change is promoted, show me whether its configuration and runtime assumptions match the target environment and block unsafe differences.

Incident responder

Must connect a production symptom to code, configuration, infrastructure, and deployment history.

Job: During an incident, rank likely causes by evidence, propose safe remediation, and preserve a complete timeline.

Agency or small-team operator

Maintains many deployments without a dedicated SRE platform.

Job: Give me a consistent inventory and safe operating workflow across projects and virtual machines without requiring enterprise infrastructure.

Canonical journey

1. Select a trusted directory.
2. Review requested scan boundaries.
3. Inspect detected repositories and services.
4. Correct or confirm low-confidence results.
5. Generate a configuration contract and local environment proposal.
6. Start and observe services.
7. compare local, staging, and production metadata.
8. validate a deployment against policy and health gates.
9. correlate a failure with code, configuration, and telemetry.
10. create an isolated, sanitized reproduction.
11. verify a fix and generate prevention artifacts.

Release 0.1 journey

Release 0.1 implements only steps 1–4:

Choose directory
  → approve scan boundary
  → discover repositories
  → classify services
  → infer tools, ports, and dependencies
  → show evidence and confidence
  → accept or correct the project map

Safety expectations

Users must always be able to answer:

• Which directory is being scanned?
• Which files were inspected?
• What fact was inferred from each file?
• Did StackCendra execute anything?
• Did any data leave the machine?
• What will change if I accept this proposal?

Release 0.1 performs read-only discovery. It does not install dependencies, execute project scripts, start containers, open network connections to discovered services, or upload repository content.

Experience requirements

• Results remain useful when AI providers are unavailable.
• Deterministic scanners produce the foundational facts.
• AI may explain or organize evidence but cannot silently replace it.
• Every inferred fact includes confidence, evidence, and correction controls.
• Rescans are incremental and preserve confirmed user corrections where valid.
• Large repositories expose progress, cancellation, exclusions, and resource limits.