dokugent dryrun¶

Simulates execution of an agent's compiled plan without running any tools or modifying any files.

This command is used to visually validate an agent’s plan file and confirm the structure of each step, including file dependencies, output expectations, and mock memory handling.

What It Does¶

• Loads the latest compiled certificate of the first detected agent in .dokugent/ops/compiled/
• Parses the plan steps
• Performs a dry run through each step:
• Logs step ID, tool used, input/output expectations, constraints, and goals
• Warns if inputs are not generated by prior steps
• Simulates reading input files (issues warning if missing)
• Generates a mock output for each step in memory (no file I/O)
• At the end, prints the full simulated memory object

When to Use¶

• After dokugent preview and before dokugent certify
• To confirm that file chaining across steps is logically sound
• To check if your tool sequence is overwriting outputs
• For basic input coverage validation (are your inputs missing or dangling?)

Example Output¶

🧩 Step: summarize_input
🔧 Tool: summarize-tool
📥 Input: input.md
📤 Output: draft-summary.md
🎯 Goal: Fulfill step summarize_input
🚧 Constraints: Must use defined tools only, Output must pass human review
📄 Input file 'input.md' simulated successfully.
🧠 Mock output generated for: draft-summary.md

Flags¶

Notes¶

• Only the first agent found in .dokugent/ops/compiled/ is simulated for now
• Future versions will allow targeting a specific agent and selecting version tags
• This command is part of Dokugent’s safe inspection loop and HITL workflow to reduce runtime errors and improve plan transparency