<objective>

Prompts

Manage prompts in the TrueFoundry Prompt Registry.

Use this skill when the user wants to:

• List prompts.
• Inspect prompt versions.
• Create a prompt.
• Update a prompt by creating a new version.
• Tag a prompt version, for example production.
• Get a prompt version FQN for Gateway usage.
• Generate code snippets or usage guidance for a prompt.

Do not create or edit agents here. Use truefoundry-agents for Agent Registry UI workflows.

</objective>

<instructions>

Preflight

1. Verify tfy login is complete. If not, use truefoundry-onboard.
2. Confirm the target tenant.
3. Confirm the ML repository or prompt repository where the prompt should live.
4. For destructive actions, use dashboard-only guidance.

UI Path

The predictable UI path is:

TrueFoundry Dashboard -> AI Gateway -> Prompt Management / Prompt Registry

Use the UI when the user wants to visually create, compare, or test prompts.

Capability Matrix

List Prompts

API fallback:

TFY_API_SH=~/.claude/skills/truefoundry-prompts/scripts/tfy-api.sh
$TFY_API_SH GET /api/ml/v1/prompts

Present:

Prompts
| Name | ID | Versions | Latest | Tags |

Create or Update Prompt

Creating a new prompt or adding a new version should be treated as a reviewed change.

Collect:

• Prompt name
• Repository / ML repo
• System and user messages
• Variables
• Model, if the prompt stores one
• Temperature / top-p / max tokens, if needed

Show the final prompt content and settings before creating or updating.

Review format before create/update:

Prompt Change
| Field | Value |
|-------|-------|
| Name | my-prompt |
| Repository | ml-repo-fqn |
| Change | create prompt / create new version |
| Variables | user_input |
| Model | model-catalog:openai:gpt-4 |
| Tags | none / production |

Then show the messages in order:

System:
...

User:
...

SDK shape:

from truefoundry.ml import ChatPromptManifest

client.prompts.create_or_update(
    manifest=ChatPromptManifest(
        name="my-prompt",
        ml_repo="ml-repo-fqn",
        messages=[
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": "{{user_input}}"},
        ],
        model_fqn="model-catalog:openai:gpt-4",
        temperature=0.7,
        max_tokens=1024,
        top_p=1.0,
    )
)

Versioning and Tags

Prompt edits create versions. Use tags such as production or staging to provide stable references.

Before applying a tag, show the current tag target if known and ask for explicit confirmation when the tag is stable or production-facing.

SDK shape:

client.prompt_versions.apply_tags(
    prompt_version_id="version-id",
    tags=["production"],
    force=True,
)

Get Prompt Version FQN

Use the UI or SDK to retrieve the FQN:

client.prompt_versions.get_by_fqn(fqn="ml-repo:prompt-name:production")

The FQN is used by Gateway prompt run flows and by TrueFoundry Agents.

Delete

Do not delete prompts or prompt versions from the agent. If deletion is requested, direct the user to the Prompt Registry UI.

</instructions>

<success_criteria>

• The user can list and inspect prompts.
• Prompt create/update is reviewed before execution.
• Prompt version tags are explained and applied only with confirmation.
• Prompt version FQNs are surfaced clearly.
• Agent authoring is routed to truefoundry-agents.

</success_criteria>