Use Cases

Use Cases

PIC is useful when AI-generated or externally collected material should be checked before it is reused, promoted, merged, routed, or treated as completed work.

LLM Output Checking

Use PIC when an LLM answer should not be trusted as completed work by default.

PIC can separate:

• the candidate claim;
• attached evidence;
• missing evidence;
• proof obligations;
• residual ledger entries;
• recommended next checks.

Coding Agent Reports

Use PIC when a coding agent reports a fix, test result, or repository finding.

PIC helps preserve what is known, what is only claimed, and what still needs verification. It does not replace tests, review, or CI.

AI Agent Workflow Verification

Use PIC inside an agent workflow when each step should leave a machine-readable trail.

PIC helps answer:

• What did the agent claim?
• Which evidence was used?
• Which verifier route is needed?
• Is the result operationally usable?
• Which obligations remain unsettled?

Phase Ecology Lab

Use Phase Lab when one report is not enough and you need to inspect a small collection of reports over time.

It can help with:

• effective packet graph inspection;
• closure candidate detection;
• execution-available path detection;
• bottleneck finding;
• queue pressure and obstruction analysis;
• reusable abstraction impact diagnostics;
• typed trace and action-boundary review;
• threshold and certificate candidates.

GitHub Actions And CI

Use PIC in CI when AI-generated output should be checked without mutating the repository.

The recommended pattern is read-only:

• run a PIC command;
• upload JSON output as an artifact;
• preserve residuals;
• do not post secrets;
• do not execute agent text.

Agent-To-Agent Messages

Use PIC when agents exchange local messages or packet candidates.

PIC can inspect message shape, provenance, identity context, nonces, and residuals. Message checks do not create broad external messaging authority.

CCR Runtime Handoff

Use PIC when a multi-agent runtime needs task and residual records that preserve the safety boundary.

PIC can emit CCR task JSONL and residual JSONL from phase plans and phase gaps. CCR or another runtime can import those records as candidate work, then apply its own leases, authority checks, provider policy, and settlement gates.

TRC Operation Candidate Review

Use PIC when a provider or real-world action needs a structured trace before any runtime decides whether to act.

TRC trace checks can show whether authority, resources, rollback, witnesses, schedule, preconditions, postconditions, and tolerance fields are present. The output is still a candidate record, not execution.

Bounded External Intake

Use PIC for explicitly supplied web, feed, repository, or local file inputs.

External content stays candidate-only until downstream evidence and verifier checks accept it. Bounded intake is not background crawling and not automatic trust.

Reusable Abstraction Checks

Use ALT features when a trace, tool pattern, or abstraction may be reusable.

PIC checks whether costs, hazards, transport, root-of-trust, telemetry, lifecycle, and baseline conditions have been counted before treating the abstraction as reusable capital.

Human Review Support

Use PIC to make review queues clearer:

• what can be inspected now;
• what is blocked;
• what needs evidence;
• what is stale;
• what should be quarantined;
• what remains residual.

PIC does not replace human judgment. It makes the state easier to inspect.