Sync kcap CLI skill with CLI v0.7.9

[w1am]

Sync kurrent-capacitor-cli skill with kcap-cli v0.7.9.

• Expand from 3 vendors (Claude/Codex/Cursor) to 8 agents, adding Copilot, Gemini, Kiro, Pi, OpenCode (SKILL.md, import, install-setup, plugins)
• Fix wrong eval-judge MCP server name (kcap-judge, not kcap-review) and list its 6 tools
• WorkOS email/SSO login path, profile add --no-probe, daemon service subcommand + --log-level + KCAP_MAX_AGENTS

Verified each change against CLI source (help-*.txt + command code).

[qodo-code-review]

PR Summary by Qodo

Sync kcap CLI skill docs with v0.7.9 (8-agent support, WorkOS login, daemon service)
📝 Documentation ✨ Enhancement 🕐 10-20 Minutes

Description

• Expand the skill/docs from 3 vendors to 8 supported agents and their install/removal flows.
• Correct judge MCP server naming (kcap-judge) and document its six read-only tools.
• Document WorkOS email/SSO login, daemon service management, and new flags/env overrides.

Diagram

graph TD
  A["SKILL.md"] --> B["install-setup.md"] --> C["plugins.md"] --> D["import.md"]
  B --> E["profiles.md"] --> F["daemon.md"]
  B --> G["sessions.md / mcp-review.md"]
  H["kcap CLI v0.7.9"] --> A

Loading

High-Level Assessment

The following are alternative approaches to this PR:

1. Generate docs from CLI help output in CI

• ➕ Prevents drift between shipped CLI flags/subcommands and SKILL.md reference files
• ➕ Makes future version bumps mostly mechanical (update fixture + regen)
• ➖ Requires a stable way to execute the CLI in CI (node/npm setup, platform differences)
• ➖ Help output formatting changes could create noisy diffs unless normalized

2. Single source-of-truth reference table + deep links

• ➕ Reduces repeated lists (agents, flags, scopes) across multiple pages
• ➕ Lowers the chance of partial updates when new agents are added
• ➖ More cross-linking overhead for readers (context not colocated)
• ➖ Still manual; doesn’t fully eliminate drift risk

Recommendation: The PR’s approach (manual sync against CLI help/code) is appropriate for a one-off v0.7.9 alignment, especially since the repo is explicitly used as a “copy names from reference files” source. If this skill will track frequent CLI releases, consider adding a lightweight generator/check in CI that compares committed reference content against kcap --help/kcap --help outputs to catch naming/flag drift early.

Files changed (8) +62 / -33

Documentation (8) +62 / -33

SKILL.mdExpand skill scope to 8 supported coding agents +3/-2

Expand skill scope to 8 supported coding agents

• Updates the skill description to reflect that kcap records sessions from eight agents (adding Copilot, Gemini, Kiro, Pi, OpenCode) instead of only Claude/Codex/Cursor. Reinforces that command/flag names must be copied from the provided references.

plugins/kurrent/skills/kurrent-capacitor-cli/SKILL.md

daemon.mdDocument daemon service subcommands, log-level flag, and KCAP_MAX_AGENTS +18/-2

Document daemon service subcommands, log-level flag, and KCAP_MAX_AGENTS

• Adds '--log-level' (and 'KCAP_DAEMON_LOG_LEVEL') to daemon start options with guidance on using debug for diagnostics. Documents 'kcap daemon service' install/status/start/stop/uninstall for per-user OS service management and adds 'KCAP_MAX_AGENTS' as a runtime override for max agents. Clarifies that 'kcap update' also refreshes opted-in agent plugins.

plugins/kurrent/skills/kurrent-capacitor-cli/references/daemon.md

import.mdAdd import coverage for Copilot/Gemini/Kiro/Pi and vendor filters +7/-1

Add import coverage for Copilot/Gemini/Kiro/Pi and vendor filters

• Expands 'kcap import' documentation to include seven historical-import sources (Copilot, Gemini, Kiro, Pi added) and notes OpenCode is live-ingest only. Documents the full set of additive vendor filter flags for importing subsets.

plugins/kurrent/skills/kurrent-capacitor-cli/references/import.md

install-setup.mdUpdate setup/login flow for WorkOS + 8-agent hook installation +15/-16

Update setup/login flow for WorkOS + 8-agent hook installation

• Updates the 'kcap setup' wizard steps to include WorkOS email/SSO vs GitHub login, optional server URL via tenant discovery, and eight-agent hook prompts. Clarifies detection rules for all agents and documents the expanded '--skip-<agent>-hooks' set for '--no-prompt' mode, including an example for “Claude only”. Updates 'kcap login'/'--discover' descriptions to reflect WorkOS discovery.

plugins/kurrent/skills/kurrent-capacitor-cli/references/install-setup.md

mcp-review.mdDescribe 'kcap mcp judge' tool surface (6 read-only tools) +1/-1

Describe 'kcap mcp judge' tool surface (6 read-only tools)

• Refines the 'kcap mcp judge' section to explicitly list the six read-only tools used to fetch session context on demand for eval/judge flows.

plugins/kurrent/skills/kurrent-capacitor-cli/references/mcp-review.md

plugins.mdAdd plugin install/remove flags and behaviors for 8 agents +15/-8

Add plugin install/remove flags and behaviors for 8 agents

• Expands 'kcap plugin install/remove' to cover Copilot, Gemini, Kiro, Pi, and OpenCode, and clarifies Claude remains the default flag-less target (no '--claude' for plugin). Documents per-agent install semantics (shared config merges vs extensions/plugins) and clarifies that only Claude/Codex support '--project' scope. Updates uninstall behavior to explicitly cover cleanup/restoration across all known agents and the expanded selective 'plugin remove' flags.

plugins/kurrent/skills/kurrent-capacitor-cli/references/plugins.md

profiles.mdDocument 'profile add --no-probe' and WorkOS-aware discovery wording +2/-2

Document 'profile add --no-probe' and WorkOS-aware discovery wording

• Adds '--no-probe' to profile creation guidance to skip server reachability checks. Updates tenant discovery description to reflect discovery via email/SSO (WorkOS) or GitHub rather than GitHub org membership only.

plugins/kurrent/skills/kurrent-capacitor-cli/references/profiles.md

sessions.mdFix eval-judge MCP server name to 'kcap-judge' (not 'kcap-review') +1/-1

Fix eval-judge MCP server name to 'kcap-judge' (not 'kcap-review')

• Corrects the MCP server name referenced by server-tagged eval judges to 'kcap-judge' and clarifies it is distinct from the plugin’s 'kcap-review' server. Keeps the overall explanation that most judges are text-only and some can fetch context via a read-only MCP surface.

plugins/kurrent/skills/kurrent-capacitor-cli/references/sessions.md

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (1) 📘 Rule violations (0) 📎 Requirement gaps (0) 🎨 UX issues (0) 🔗 Cross-repo conflicts (0) 📜 Skill insights (0)

1. Evals outdated vs new agents ✓ Resolved 📘 Rule violation ☼ Reliability R4 / Evals

Description

The skill description and setup docs were updated to cover 8 agents and email/SSO (WorkOS) login,
but the existing eval cases still assert the older GitHub-only login flow and only 3 supported
agents/older skip flags. This violates the requirement to update eval coverage whenever skill
content/description changes, and will cause quality/trigger regression to go unmeasured.

Code

plugins/kurrent/skills/kurrent-capacitor-cli/SKILL.md[R4-7]

  Use when installing, configuring, or operating the kcap CLI, aka the
  Kurrent Capacitor CLI / Capacitor CLI (npm @kurrent/kcap), which records
-  Claude Code, Codex, and Cursor sessions to a Capacitor server. Covers
+  coding-agent sessions (Claude Code, Codex, Cursor, Copilot, Gemini, Kiro,
+  Pi, OpenCode) to a Capacitor server. Covers

Evidence

PR content updates the skill to cover eight agents and the WorkOS email/SSO login path, but
evals/quality/cases/kurrent-capacitor-cli.yaml still requires GitHub-only login language and only
names Claude/Codex/Cursor with the older opt-out flags. This mismatch indicates eval coverage was
not updated alongside the skill changes, failing PR Compliance ID 15.

Rule R4 / Evals: Evals coverage must be updated when skill content or descriptions change (quality vs trigger axes)
plugins/kurrent/skills/kurrent-capacitor-cli/SKILL.md[4-12]
plugins/kurrent/skills/kurrent-capacitor-cli/references/install-setup.md[54-98]
evals/quality/cases/kurrent-capacitor-cli.yaml[8-29]
evals/quality/cases/kurrent-capacitor-cli.yaml[38-42]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
Skill docs now describe 8 supported agents and WorkOS email/SSO login, but eval artifacts still encode the older behavior (GitHub-only login, 3 agents, old `--skip-*-hooks` set). Update evals so they match the new documented behavior and add/adjust negative trigger cases as needed.

## Issue Context
This PR expands supported agents and changes login/setup wording; without eval updates, routing/quality checks will validate against stale expectations.

## Fix Focus Areas
- evals/quality/cases/kurrent-capacitor-cli.yaml[8-42]
- evals/trigger/kurrent-capacitor-cli.json[1-92]
- plugins/kurrent/skills/kurrent-capacitor-cli/references/install-setup.md[54-98]
- plugins/kurrent/skills/kurrent-capacitor-cli/SKILL.md[4-12]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

2. Onboarding login summary stale ✓ Resolved 🐞 Bug ⚙ Maintainability

Description

SKILL.md still summarizes kcap setup as performing “GitHub login”, but the updated setup wizard
documentation now supports email/SSO (WorkOS) discovery login as well. This inconsistency can cause
the skill router to give outdated onboarding/auth guidance.

Code

plugins/kurrent/skills/kurrent-capacitor-cli/references/install-setup.md[R56-60]

+1. **Server URL** — provided by your admin, e.g. `https://my-tenant.kcap.ai`. Can be skipped when the server supports tenant discovery (see Login).
+2. **Login** — at discovery you choose how to sign in: **email / SSO (WorkOS)** or **GitHub** (browser/PKCE by default, falls back to Device Flow). When the server's auth provider is `None`, setup prints "no login required" and skips this step.
3. **Default visibility** — `private` / `org_public` / `public`; the dashboard-visibility default applied to new sessions.
-4. **Coding-agent hooks** — one yes/no per detected agent (Claude Code / Codex CLI / Cursor).
+4. **Coding-agent hooks** — one yes/no per detected agent (Claude Code / Codex CLI / Cursor / Copilot CLI / Gemini CLI / Kiro CLI / Pi / OpenCode).
5. **Daemon name**.

Evidence

SKILL.md still says the onboarding wizard includes “GitHub login”, while the updated setup
reference explicitly states discovery login can be email/SSO (WorkOS) or GitHub.

plugins/kurrent/skills/kurrent-capacitor-cli/SKILL.md[23-27]
plugins/kurrent/skills/kurrent-capacitor-cli/references/install-setup.md[54-60]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
The skill router summary in `SKILL.md` says setup includes “GitHub login”, but the detailed setup reference now documents discovery login via email/SSO (WorkOS) or GitHub.

### Issue Context
`SKILL.md` is loaded as standing instruction and used for routing/summary; it should not contradict `references/install-setup.md`.

### Fix Focus Areas
- plugins/kurrent/skills/kurrent-capacitor-cli/SKILL.md[23-27]
- plugins/kurrent/skills/kurrent-capacitor-cli/references/install-setup.md[54-60]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

3. OS-service guidance in references ✓ Resolved 📘 Rule violation ⚙ Maintainability R1-R2

Description

New content in references/daemon.md adds prescriptive operational instructions (behavioral
guidance) rather than purely descriptive facts. This violates the R1/R2 layering requirement that
references contain descriptive facts only and that behavioral guidance/guards live in SKILL.md.

Code

plugins/kurrent/skills/kurrent-capacitor-cli/references/daemon.md[R22-35]

+## Run as an OS service
+
+For a daemon that survives logout, auto-restarts on crash/`SIGKILL`, and starts at login, register it as a per-user OS service (launchd on macOS, systemd user unit on Linux, Scheduled Task on Windows) instead of running `daemon start -d` by hand.
+
+```bash
+kcap daemon service install                 # register + start; pins profile, captures PATH
+kcap daemon service install --name laptop --profile work --max-agents 10 --no-start
+kcap daemon service status                  # installed / running state
+kcap daemon service start                   # start an installed (stopped) service
+kcap daemon service stop                    # stop but keep installed
+kcap daemon service uninstall               # stop and remove the unit
+```
+
+`install` pins the profile via `KCAP_PROFILE` and captures your shell `PATH`, so `claude` / `codex` resolve the same way they do interactively. `--no-start` registers without starting. All actions take `--name N` to target a specific daemon name.

Evidence

The added section explicitly instructs users to register the daemon as a per-user OS service
"instead of" running daemon start -d manually, which is behavioral guidance embedded in a
reference file. PR Compliance ID 7 requires references to be descriptive facts only, with behavioral
guidance/guards placed in SKILL.md.

Rule R1-R2: Content must follow R1/R2 layering: section ownership and correct placement of facts vs instructions/guards
plugins/kurrent/skills/kurrent-capacitor-cli/references/daemon.md[22-35]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`references/daemon.md` includes newly-added prescriptive guidance about how users "should" run the daemon (register as an OS service instead of running detached manually). Per the layering rules, references should be descriptive/factual, while behavioral guidance/guards belong in SKILL.md.

## Issue Context
This section was added in the PR and is actionable instruction, not just command reference.

## Fix Focus Areas
- plugins/kurrent/skills/kurrent-capacitor-cli/references/daemon.md[22-35]
- plugins/kurrent/skills/kurrent-capacitor-cli/SKILL.md[19-52]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

4. Wrong vendor-flag count 🐞 Bug ≡ Correctness

Description

references/plugins.md claims “Eight agents have a vendor flag” but the listed flags are seven
agent-specific flags plus --skills, which is explicitly agent-agnostic. This can mislead users
(and the skill agent) about the actual kcap plugin flag surface.

Code

plugins/kurrent/skills/kurrent-capacitor-cli/references/plugins.md[10]

+Eight agents have a vendor flag (`--codex`, `--cursor`, `--copilot`, `--gemini`, `--kiro`, `--pi`, `--opencode`, plus `--skills` for the agent-agnostic skills). Claude is the **default, flag-less** target — there is no `--claude` on `plugin` (it only exists as an `import` vendor filter).

Evidence

The command synopsis lists seven agent flags (--codex, --cursor, --copilot, --gemini,
--kiro, --pi, --opencode) and separately includes --skills; the following sentence
incorrectly counts --skills as part of the agent/vendor-flag count.

plugins/kurrent/skills/kurrent-capacitor-cli/references/plugins.md[5-11]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
`references/plugins.md` incorrectly states that eight agents have a vendor flag while counting `--skills`, which is not an agent/vendor flag.

### Issue Context
This file is used as the source-of-truth reference for the kcap CLI skill; inaccurate flag surface descriptions degrade answer quality.

### Fix Focus Areas
- plugins/kurrent/skills/kurrent-capacitor-cli/references/plugins.md[5-11]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---