Background
The post-v0.4.6 report work now exposes richer local evidence across JSON, HTML, and Markdown overview formats: incident timelines and tool authority summaries are visible in shareable artifacts. The default text overview remains much thinner, even though it is the fastest terminal path for a user running agenttrace --overview -f text or reading command output in a terminal.
Evidence
Validated on master b93c7ba after #210 and #212:
agenttrace --demo --overview -f json exposes incident_timelines, summary.tool_authority, and surfaces.authority_categories.agenttrace --demo --overview -f html renders Incident timeline and Tool authority.agenttrace --demo --overview -f markdown renders Incident timeline and Tool authority.agenttrace --demo --overview -f text currently renders totals, By Agent, By Model, and Recent Anomalies only. It does not mention incident timeline cues, touched surface cues, highest authority category, authority counts, or high-authority tools.
User value
A terminal-first user should get enough evidence from the default readable text output to understand why a run is risky without switching to JSON/Markdown/HTML. The text view does not need full tables, but it should carry the most actionable cues that already exist in other report formats.
Adoption rationale
The first CLI experience should demonstrate the product's current value: not just cost and anomalies, but local debugging evidence such as failure loops, idle gaps, touched tools, and tool authority. A compact text summary helps users see the value before they export a report.
Suggested scope
Add compact, terminal-friendly sections or lines to text overview output when data exists:
- a short incident/timeline evidence block with the top few high-signal items;
- a short tool authority block with highest category and high-authority tools or category counts;
- keep the output readable and bounded.
Add or extend a repeatable check so demo text output cannot silently miss these cues.
Non-goals
- Do not make text output as dense as Markdown/HTML tables.
- Do not change JSON field names or authority classification rules.
- Do not add hosted telemetry, upload, or external sharing behavior.
- Do not change release, package, npm, Homebrew, or public site surfaces.
Acceptance criteria
agenttrace --demo --overview -f text includes visible incident/timeline evidence when available.agenttrace --demo --overview -f text includes visible tool authority evidence when available.- Existing JSON, Markdown, and HTML overview contracts remain compatible.
- A CI/script check covers the new text cues.
go test ./..., go build, scripts/ci/check-report-semantics.sh, and relevant output/docs checks pass if touched.- Copy remains local-first and does not imply uploaded logs or hosted tracing.
Suggested lane
lane/product, with Quality help for the regression check if useful.
Risk
Medium-low. This changes terminal output, so the main risk is making the default text overview noisy. Keep it compact and evidence-first.
Source
Product dogfood after #210/#212 on master b93c7ba.
Background
The post-v0.4.6 report work now exposes richer local evidence across JSON, HTML, and Markdown overview formats: incident timelines and tool authority summaries are visible in shareable artifacts. The default text overview remains much thinner, even though it is the fastest terminal path for a user running
agenttrace --overview -f textor reading command output in a terminal.Evidence
Validated on master
b93c7baafter #210 and #212:agenttrace --demo --overview -f jsonexposesincident_timelines,summary.tool_authority, andsurfaces.authority_categories.agenttrace --demo --overview -f htmlrendersIncident timelineandTool authority.agenttrace --demo --overview -f markdownrendersIncident timelineandTool authority.agenttrace --demo --overview -f textcurrently renders totals, By Agent, By Model, and Recent Anomalies only. It does not mention incident timeline cues, touched surface cues, highest authority category, authority counts, or high-authority tools.User value
A terminal-first user should get enough evidence from the default readable text output to understand why a run is risky without switching to JSON/Markdown/HTML. The text view does not need full tables, but it should carry the most actionable cues that already exist in other report formats.
Adoption rationale
The first CLI experience should demonstrate the product's current value: not just cost and anomalies, but local debugging evidence such as failure loops, idle gaps, touched tools, and tool authority. A compact text summary helps users see the value before they export a report.
Suggested scope
Add compact, terminal-friendly sections or lines to text overview output when data exists:
Add or extend a repeatable check so demo text output cannot silently miss these cues.
Non-goals
Acceptance criteria
agenttrace --demo --overview -f textincludes visible incident/timeline evidence when available.agenttrace --demo --overview -f textincludes visible tool authority evidence when available.go test ./...,go build,scripts/ci/check-report-semantics.sh, and relevant output/docs checks pass if touched.Suggested lane
lane/product, with Quality help for the regression check if useful.
Risk
Medium-low. This changes terminal output, so the main risk is making the default text overview noisy. Keep it compact and evidence-first.
Source
Product dogfood after #210/#212 on master
b93c7ba.