Improve CLI human output UX

[odyslam]

Summary

• Make default CLI output human-first across command surfaces, including auth, config, workflow, schema, API discovery, dry-run, export, job, artifact, request log, collection, incident, and raw API response views.
• Add --toon as the compact agent-readable output mode while preserving --json and hidden legacy --format compatibility.
• Polish the default human UX across the audited command set: source metadata now says Source: Phylax platform index, empty lists say what was not found, mixed search results show all buckets, project/account/deployment/member outputs get dedicated layouts, and human errors include API-provided reasons without raw envelope instructions.
• Update agent-facing docs, the LLM guide, and smoke checks to use --toon.
• Bump the workspace patch version from 1.4.2 to 1.4.3.

Reviewer Guide

• Inspect crates/pcl/common/src/args.rs for the three output modes: default human, --toon, and --json.
• Inspect crates/pcl/core/src/api.rs for human renderers, next-action filtering, search/project/account/deployment/member formatting, source labels, and cleaned error rendering.
• Inspect crates/pcl/core/src/config.rs and crates/pcl/core/src/surface.rs for safer human next actions around config and doctor failures.
• Run cargo run --locked -q -p pcl -- incidents --limit 3 and confirm the default view is a readable table/summary, not the full envelope dump.
• Also spot-check search --query linea, projects --project-id <id>, releases --project <id>, access --pending, integrations --project <id> --provider slack, api call GET /views/public/incidents --query limit=3 --dry-run, and representative errors like api list --limit 3.
• Compare --toon and --json on representative commands such as workflows, schema list, api list --filter incidents, auth status, and requests list --limit 3.
• Check AGENTS.md, README.md, and scripts/agent-smoke.sh for the new agent-facing --toon guidance.

Validation

• make ci
• make audit
• cargo run --locked -q -p pcl -- --version -> pcl 1.4.3
• Focused human-output regression tests: cargo test -p pcl-core --lib 'human_' -- --nocapture
• Default-output audit over 928 lines in /tmp/pcl-human-full-audit-final.txt found no legacy/bad UX markers: Details:, Schema: pcl.envelope.v1, Request:, Response:, from offchain, item(s), raw clap error:/Usage:, config delete, Read error.http.body, Use data.*, or blank Body:.

[odyslam]

Follow-up commit 6fdea9b completes the broader default-output audit pass.

What changed in this commit:

• Replaced internal collection metadata like Fetched ... from offchain with human labels: Updated: and Source: Phylax platform index.
• Added targeted human layouts for project detail, account detail, deployments, integrations, protocol manager state, transfer state, members, invitations, releases, events, and mixed search results.
• Cleaned successful empty states from generic No results. into concrete text like No assertions found. / No releases found. and removed placeholder next steps when there is no item to inspect.
• Cleaned human errors so clap parse failures no longer dump raw error:/Usage: blocks, and API failures surface the API-provided reason plus request ID.
• Accepted uppercase HTTP methods for raw API calls (GET, POST, etc.) because humans type them that way.
• Updated config/doctor next actions to avoid destructive or irrelevant suggestions in healthy/default human flows.

Validation after this commit:

• make ci
• make audit
• cargo test -p pcl-core --lib 'human_' -- --nocapture
• 928-line safe-command/default-error audit in /tmp/pcl-human-full-audit-final.txt with no old envelope/bad UX markers.

[odyslam]

Follow-up commit bef3189 fixes the user-project discovery path:

• Makes pcl projects --mine the primary documented flag and keeps --home as an alias.
• Adds a dedicated default human renderer for /views/projects/home, so it prints a Your projects table with project, slug, network, visibility, and ID instead of backend field names.
• Updates next actions and README guidance to point people and agents at pcl projects --mine.
• Adds parser/request/renderer regression tests for --mine, --home, and the human project-home output.

Local validation after this commit:

• cargo test -p pcl-core
• cargo build -p pcl
• ./target/debug/pcl projects --help
• ./target/debug/pcl projects --mine
• ./target/debug/pcl projects --home --dry-run
• make ci
• make audit

[odyslam]

Follow-up commit 1450743 is the command-surface DevEx audit pass.

I ran the default command surface and representative read-only/error paths locally, then fixed the remaining rough spots:

• pcl search no longer sends a bad /search request with no query; it fails locally with useful examples.
• Project-scoped commands now point users to pcl projects --mine and the relevant command help when --project / --project-id is missing.
• pcl events now uses the same human workflow error path as the other API workflows instead of raw clap required-argument text.
• Missing subcommands for api, auth, config, and export now say what to do next instead of only echoing the command description.
• Unknown/invalid subcommands now get better next actions, including schema get / workflows show hints for common mistakes.
• apply, build, and download runtime failures now use typed human error envelopes instead of Code: unknown.
• Auth/whoami next actions now send users to pcl projects --mine.
• Deployment/project count labels now say 0 contracts / 0 assertions instead of generic item wording.

Validation:

• make ci
• make audit
• final local audit logs: 469 no-flag/default lines, 310 representative workflow lines, and 117 error-path lines scanned with no legacy markers like Details:, Schema:, raw Request:/Response:, from offchain, item(s), raw clap Usage:, Code: unknown, or accidental Use data.* guidance.

[odyslam]

Closing this oversized PR in favor of smaller stacked PRs. The current diff is too large to review cleanly; I am preserving the work and will reopen it as a reviewable stack with the generated-client pruning included in the final layer.