Append System Prompt

Steer any agent with a one-off system-prompt suffix — no file or YAML edits

--append-system-prompt appends an arbitrary text suffix to the assembled system prompt for a single invocation — never persisted, and applied at the very end so the cacheable prompt prefix stays intact.

Quick Start

1. Append literal text

   praisonai code --append-system-prompt "Always answer in French"

2. Read the suffix from a file

   praisonai run "Refactor this module" --append-system-prompt @policy.md

3. Set the env var (CI-friendly)

   export PRAISONAI_APPEND_SYSTEM_PROMPT="Be brief. No preamble."
   praisonai chat

How It Works

The CLI resolves the flag value, exports it as PRAISONAI_APPEND_SYSTEM_PROMPT, and the core Agent reads that single env var when it assembles its system prompt. Because every construction path — interactive TUI, one-shot prompt, YAML agents, and Python Agent(...) — funnels through the same env var, the suffix reaches all of them without a new constructor parameter.

The suffix is appended after the stable prompt prefix (and after any path-scoped glob rules), so the cacheable base prompt is preserved and prompt-cache reuse is unaffected. A whitespace-only or unset value is a no-op, and the suffix is never written to any agent definition, YAML, or session backstory.

Choosing a Form

Pick the form that matches how long the guidance should live and how it’s supplied.

When the value starts with @, the file contents are read (a missing file falls back to treating the raw value as literal text, so a run is never derailed).

Where It Applies

One flag, three commands, same value shape everywhere. Exporting the env var also reaches YAML and Python agents in the same process.

Common Patterns

Steer one CLI run in French, no code changes:

praisonai code --append-system-prompt "Always answer in French"

Read the suffix from a policy file:

praisonai run "Refactor this module" --append-system-prompt @policy.md

Set once for a CI job, use across every command:

export PRAISONAI_APPEND_SYSTEM_PROMPT="Reply in strict JSON, no prose."
praisonai chat

Same knob from Python — every Agent constructed after the export picks it up:

import os
from praisonaiagents import Agent

os.environ["PRAISONAI_APPEND_SYSTEM_PROMPT"] = "Reply concisely; no bullet points."
agent = Agent(instructions="You are a research assistant")
agent.start("Summarise transformer attention in 3 lines.")

Best Practices

Related