| name | sc-bootstrap |
| description | Verify the Shadow Clone install is complete for Kimi, and walk the user through completing it if not |
| type | prompt |
| whenToUse | When the user wants to check or finish their Shadow Clone installation for use with Kimi |
You are bootstrapping a Shadow Clone install for Kimi. The skills give Kimi the
/skill:sc-* surface, but the review loop also needs the bridge scripts, the
protocol library, and the reviewer persona on disk — those come from the source
clone's bridge/install.sh and kimi/install.sh, not from the skills alone.
Your job is to detect what's installed and tell the user the smallest
next step to complete the install.
Step 1 — Detect install state
Run this exact check via the Bash tool:
echo "BRIDGE_OK=$([ -x ~/.claude/sc/ask-glm.sh ] && [ -x ~/.claude/sc/ask-claude.sh ] && [ -x ~/.claude/sc/ask-grok.sh ] && [ -x ~/.claude/sc/ask-kimi.sh ] && echo yes || echo no)"
echo "PROTOCOLS_OK=$([ -d ~/.claude/sc/protocols ] && [ "$(ls ~/.claude/sc/protocols 2>/dev/null | wc -l)" -ge 10 ] && echo yes || echo no)"
echo "REVIEWER_OK=$([ -f ~/.config/opencode/agent/sc-echo-reviewer.md ] && echo yes || echo no)"
echo "KIMI_SKILLS_OK=$([ -f "${KIMI_CODE_HOME:-$HOME/.kimi-code}/skills/sc-echo/SKILL.md" ] && echo yes || echo no)"
echo "KIMI_SKILLS_DIR=${KIMI_CODE_HOME:-$HOME/.kimi-code}/skills"
echo "REVIEWER_CLI=$(command -v claude >/dev/null 2>&1 && echo claude || { command -v opencode >/dev/null 2>&1 && echo opencode || { command -v grok >/dev/null 2>&1 && echo grok || { command -v kimi >/dev/null 2>&1 && echo kimi || echo none; }; }; })"
Parse the six KEY=value lines into a state map.
Step 2 — Decide the next step
Evaluate the cases in the order listed (A → B → C → D → E) and emit the
first message whose conditions match — the cases can overlap (e.g. B and D),
so this precedence is what makes the guidance deterministic. Substitute the
detected values where shown, then stop. Do not chain into another command.
A. Fully installed (BRIDGE_OK=yes, PROTOCOLS_OK=yes, REVIEWER_OK=yes, KIMI_SKILLS_OK=yes, REVIEWER_CLI≠none)
✅ Shadow Clone is fully installed for Kimi.
- Bridge scripts at
~/.claude/sc/ — ready
- Canonical protocols deployed — ready
- Echo reviewer persona registered — ready
- Kimi skills at
<KIMI_SKILLS_DIR>/sc-*/ — ready
- Reviewer CLI available:
<REVIEWER_CLI> — ready
You can use /skill:sc-echo to turn on paired-review for the next work unit,
or /skill:sc-help for the skill catalog.
B. Skills present, bridge missing (KIMI_SKILLS_OK=yes, BRIDGE_OK=no OR PROTOCOLS_OK=no)
⚠ The Kimi skills are installed, but the bridge scripts, protocols, or reviewer
persona are NOT on disk. /skill:sc-echo (paired review) DOES NOT WORK without
them.
To complete the install, clone the repo and run both installers:
git clone --depth 1 https://github.com/Ignis-AI-Labs/shadow-clone.git
cd shadow-clone
bash bridge/install.sh
bash kimi/install.sh
bash scripts/sc-doctor.sh
(If REVIEWER_CLI was also none, you still need a reviewer CLI afterwards —
see case D.)
C. Bridge present, skills missing (BRIDGE_OK=yes, KIMI_SKILLS_OK=no)
⚠ The Shadow Clone bridge is installed, but the Kimi skills are not on disk.
From your clone of the repo, run:
bash kimi/install.sh
bash scripts/sc-doctor.sh
(If REVIEWER_CLI was also none, you still need a reviewer CLI afterwards —
see case D.)
D. No reviewer CLI (REVIEWER_CLI=none)
⚠ The review loop needs a reviewer CLI on PATH — claude (default when Kimi
builds), opencode, grok, or kimi (self-review). None was found. Install
one, then re-run /skill:sc-bootstrap. Paired review works best when a
different model holds the reviewer pen — self-review (/skill:sc-echo kimi)
is the fallback, not the ideal.
E. Mixed state (any other combination)
⚠ Shadow Clone is partially installed. Re-run the installers from your clone
of the repo to refresh everything:
cd /path/to/your/shadow-clone-clone
bash bridge/install.sh
bash kimi/install.sh
bash scripts/sc-doctor.sh
Detected state:
- BRIDGE_OK=
- PROTOCOLS_OK=
- REVIEWER_OK=
- KIMI_SKILLS_OK=
- REVIEWER_CLI=
Step 3 — Stop
After emitting the matching message, stop. Do not run additional commands
on the user's behalf. The user runs the install themselves so they
understand what landed where.