| name | ingest-system-prompt |
| model | sonnet |
| description | Produces a stored MemoryEntry (structured sections + ADA disposition indicators) and one or more MemorySeeds by parsing a system prompt from an external AI tool into Dojo memory. Use when: "ingest this system prompt", "store this prompt in memory", "parse this agent's rules", "import this Cursor prompt", "analyze what this prompt does". |
| license | proprietary |
| category | wisdom-garden |
| triggers | ["ingest this system prompt","store this prompt in memory","parse this agent's rules","analyze what this prompt does"] |
| tier | 1 |
| agents | ["primary"] |
| tool_dependencies | ["file_system"] |
| inputs | [{"name":"system_prompt","type":"string","description":"The external AI system prompt to parse and store in Dojo memory","required":true}] |
| outputs | [{"name":"memory_entry","type":"string","description":"Stored MemoryEntry with structured sections and ADA disposition indicators, plus one or more MemorySeeds"}] |
I. Philosophy
A system prompt is not just configuration — it is a behavioral contract. Every
rule, constraint, and instruction in a system prompt encodes assumptions about
pace, depth, tone, initiative, and trust. Ingesting it properly means
converting that implicit behavioral contract into explicit, queryable knowledge
that the Dojo agent can reason about and apply.
The goal is not to replicate the foreign prompt verbatim. It is to extract the
shape of the intended agent behavior and store it in a form the Dojo memory
system can cross-reference against current dispositions.
II. When to Use
Use this skill when:
- A user imports a system prompt from Cursor, Windsurf, GitHub Copilot, Cline,
Continue, or any other AI coding or writing tool.
- An organization wants to audit what behavioral rules are governing a
third-party AI instance.
- A developer is migrating from one AI tool to Dojo and needs to preserve the
intended agent behavior.
- A team wants to compare the behavioral profile of two different system prompts.
- A system prompt needs to be version-controlled in the Dojo memory store.
Do not use this skill to execute or simulate the foreign system prompt. Its
purpose is ingestion and analysis only.
III. Workflow
Step 1 — Read the system prompt file.
Accept either a file path or raw text. If a file path is given, read its
contents. If the source tool is not stated explicitly, attempt to identify it
from the file header, filename convention (e.g., .cursorrules,
copilot-instructions.md), or ask the user before proceeding.
Step 2 — Identify the source tool.
Determine the originating tool. This becomes the source_tool metadata value.
If the version or date is available (from filename, header, or git history),
capture it as the version value. Use "unknown" if not determinable.
Step 3 — Parse into structured sections.
Scan the prompt for structural markers: markdown headings, numbered lists,
labeled blocks (e.g., ## Rules, ## Constraints, <!-- TOOLS -->), or
implicit paragraph groupings. Map content into the following canonical sections:
core_instructions — the primary task or role definitionbehavioral_rules — explicit dos and don'tstool_definitions — any tool names, function signatures, or MCP endpoints
defined within the prompt- — hard limits (token caps, forbidden outputs, format rules)
