Add CodeFlicker (KwaiPilot) system prompts and tools#412
Conversation
WalkthroughNew prompt specification files added for CodeFlicker AI agent modes (Browser, Review, Discuss, Duet, Jam, Plan, Preview, Research sub-agents), tool documentation, memory system, and review templates—11 files totaling approximately 1,890 lines of agent behavior definitions, workflow constraints, and operational guidelines. Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@CodeFlicker/Agent` Prompt (Browser SubAgent).txt:
- Around line 6-10: Replace the two blank placeholders on the instruction lines
with a clear trigger phrase so the agent detects decomposition scenarios;
specifically, change the first blank to "multi-page/multi-domain/multi-source"
so the line reads "When you encounter a multi-page/multi-domain/multi-source
scenario, provide a clear explanation...", and change the second blank to the
same phrase so the line reads "Important: multi-page/multi-domain/multi-source
is NOT an error - it's a request for task decomposition..."; this ensures the
decomposition trigger references the exact scenario types discussed earlier.
In `@CodeFlicker/Agent` Prompt (Jam Mode).txt:
- Line 54: Complete the incomplete citation instruction in "Agent Prompt (Jam
Mode).txt" by adding the full format specification and explanation used in other
prompts (e.g., the example in "Cursor Prompts/Chat Prompt.txt"); specifically,
append a code block example showing the required literal format
```startLine:endLine:filepath followed by the cited code, and a short sentence
clarifying that this is the ONLY acceptable format and that startLine and
endLine are numeric line numbers corresponding to the cited region. Update the
line that currently ends with "You MUST use the following format when citing
code regions or blocks: " to include that example block and the explanatory
sentence so the instruction is complete and consistent with other prompts.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: 588a281b-6530-468c-8848-a4d84b38c5f6
📒 Files selected for processing (11)
CodeFlicker/Agent Prompt (Browser SubAgent).txtCodeFlicker/Agent Prompt (Code Review).txtCodeFlicker/Agent Prompt (Discuss Mode).txtCodeFlicker/Agent Prompt (Duet Mode).txtCodeFlicker/Agent Prompt (Jam Mode).txtCodeFlicker/Agent Prompt (Plan Mode).txtCodeFlicker/Agent Prompt (Preview SubAgent).txtCodeFlicker/Agent Prompt (Research SubAgent).txtCodeFlicker/Agent Tools.txtCodeFlicker/Memory System Prompt.txtCodeFlicker/Review Report Templates.txt
| - When you encounter a scenario, provide a clear explanation of why the task is too large | ||
| - List the specific pages/domains that would need to be visited | ||
| - Suggest logical sub-tasks that could be handled separately | ||
| - **Example**: This task requires visiting 4 different pages across 2 domains. Suggested breakdown: 1) Page A to B, 2) Page C to D. Please decompose into smaller sub-tasks. | ||
| - **Important**: is NOT an error - it's a request for task decomposition to ensure successful execution |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Validate whether blank placeholders are isolated or recurring extraction artifacts.
set -euo pipefail
echo "== Inspect suspicious lines with context =="
rg -n -C2 "encounter a scenario|Important.* is NOT an error" "CodeFlicker/Agent Prompt (Browser SubAgent).txt"
echo
echo "== Check for similar blank-double-space placeholder patterns in CodeFlicker prompts =="
rg -n -P "\b[a-zA-Z]+\s{2,}[a-zA-Z]+\b|:\s{2,}\w|-\s\*\*Important\*\*:\s\s" CodeFlickerRepository: x1xhlol/system-prompts-and-models-of-ai-tools
Length of output: 4584
Task decomposition trigger is incomplete — blanks prevent agent from recognizing when to decompose
Lines 6 and 10 contain blank placeholders that make the decomposition trigger under-specified. Context from lines 4–5 (discussing "Multi-Page Tasks" and "Multi-Source Data") suggests the blanks should reference either "multi-page", "multi-domain", or "multi-source" scenarios. Without these terms, the agent cannot reliably recognize when decomposition is required.
Fill in both blanks with the appropriate term(s) so the instruction reads: When you encounter a [term] scenario... and [term] is NOT an error...
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@CodeFlicker/Agent` Prompt (Browser SubAgent).txt around lines 6 - 10, Replace
the two blank placeholders on the instruction lines with a clear trigger phrase
so the agent detects decomposition scenarios; specifically, change the first
blank to "multi-page/multi-domain/multi-source" so the line reads "When you
encounter a multi-page/multi-domain/multi-source scenario, provide a clear
explanation...", and change the second blank to the same phrase so the line
reads "Important: multi-page/multi-domain/multi-source is NOT an error - it's a
request for task decomposition..."; this ensures the decomposition trigger
references the exact scenario types discussed earlier.
| 3. Use <br/> for line breaks within node labels, not actual line breaks. | ||
| 4. Labels should always be surrounded by double quotes ("") so that it doesn't create any syntax errors if there are special characters inside. | ||
| 5. Finally once it seems like you've reached a good plan, then you can make code changes. </mermaid> | ||
| You MUST use the following format when citing code regions or blocks: No newline at end of file |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Verify whether Jam Mode prompt is intentionally ending at the code fence
# and compare with nearby prompt artifacts for similar pattern continuity.
set -euo pipefail
echo "== Locate file =="
fd -a "Agent Prompt (Jam Mode).txt" CodeFlicker
echo
echo "== Tail with line numbers =="
nl -ba "CodeFlicker/Agent Prompt (Jam Mode).txt" | tail -n 40
echo
echo "== Search for the same citation-format phrase across CodeFlicker prompts =="
rg -n -C2 "You MUST use the following format when citing code regions or blocks|citing code regions or blocks" CodeFlicker
echo
echo "== Check for unmatched fenced code blocks in this file =="
python - << 'PY'
from pathlib import Path
p = Path("CodeFlicker/Agent Prompt (Jam Mode).txt")
text = p.read_text(encoding="utf-8", errors="replace")
count = text.count("```")
print("Fence count:", count, "(odd => likely unmatched)")
PYRepository: x1xhlol/system-prompts-and-models-of-ai-tools
Length of output: 197
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "== Search for Jam Mode prompt file =="
find . -type f -name "*Jam Mode*" 2>/dev/null || echo "No match with direct find"
echo
echo "== List CodeFlicker directory =="
ls -la CodeFlicker/ 2>/dev/null | head -20 || echo "CodeFlicker directory not found"
echo
echo "== Check git tracked files in CodeFlicker =="
git ls-files CodeFlicker/ 2>/dev/null | grep -i "jam\|prompt" || echo "No jam/prompt files found"
echo
echo "== Get all files in CodeFlicker =="
git ls-files CodeFlicker/ 2>/dev/null | head -20
echo
echo "== Search for citation format phrase =="
rg -i "you must use the following format" . 2>/dev/null || echo "Pattern not found"Repository: x1xhlol/system-prompts-and-models-of-ai-tools
Length of output: 2789
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "== Total line count in Jam Mode file =="
wc -l "CodeFlicker/Agent Prompt (Jam Mode).txt"
echo
echo "== Content around line 54 and beyond (last 60 lines) =="
sed -n '50,$p' "CodeFlicker/Agent Prompt (Jam Mode).txt" | cat -n
echo
echo "== Count backticks in Jam Mode file =="
grep -o '```' "CodeFlicker/Agent Prompt (Jam Mode).txt" | wc -l
echo
echo "== Compare: View same phrase in Cursor Chat Prompt =="
echo "--- Lines around citation format phrase in Chat Prompt ---"
grep -n -A 5 "You MUST use the following format when citing code regions or blocks" "./Cursor Prompts/Chat Prompt.txt" | head -20
echo
echo "== Count backticks in Chat Prompt =="
grep -o '```' "./Cursor Prompts/Chat Prompt.txt" | wc -lRepository: x1xhlol/system-prompts-and-models-of-ai-tools
Length of output: 866
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "== Context around citation phrase in Chat Prompt (20 lines after) =="
grep -n -A 20 "You MUST use the following format when citing code regions or blocks" "./Cursor Prompts/Chat Prompt.txt" | head -30
echo
echo "== Full Jam Mode content for inspection =="
cat "CodeFlicker/Agent Prompt (Jam Mode).txt"Repository: x1xhlol/system-prompts-and-models-of-ai-tools
Length of output: 9318
Complete the missing citation format specification at end of file
Line 53 ends with You MUST use the following format when citing code regions or blocks: but provides no format specification or example. Compare with other prompts (e.g., Cursor Prompts/Chat Prompt.txt) which show the complete instruction:
Expected format (from Chat Prompt)
You MUST use the following format when citing code regions or blocks:
```startLine:endLine:filepath
// ... existing code ...
This is the ONLY acceptable format for code citations. The format is ```startLine:endLine:filepath where startLine and endLine are line numbers.
</details>
Add the format specification and explanation to complete this instruction.
<details>
<summary>🤖 Prompt for AI Agents</summary>
Verify each finding against the current code and only fix it if needed.
In @CodeFlicker/Agent Prompt (Jam Mode).txt at line 54, Complete the incomplete
citation instruction in "Agent Prompt (Jam Mode).txt" by adding the full format
specification and explanation used in other prompts (e.g., the example in
"Cursor Prompts/Chat Prompt.txt"); specifically, append a code block example
showing the required literal format ```startLine:endLine:filepath followed by
the cited code, and a short sentence clarifying that this is the ONLY acceptable
format and that startLine and endLine are numeric line numbers corresponding to
the cited region. Update the line that currently ends with "You MUST use the
following format when citing code regions or blocks: " to include that example
block and the explanatory sentence so the instruction is complete and consistent
with other prompts.
</details>
<!-- fingerprinting:phantom:triton:hawk -->
<!-- This is an auto-generated comment by CodeRabbit -->
1 participant
Add CodeFlicker (KwaiPilot) system prompts and tools.
Some of them are in Chinese because they were that way from the beginning.
Summary by CodeRabbit