MCP reference / Prompts
Prompts
TAM's MCP prompts: 13 automation scenarios for planning, triage, sprint operations, and release readiness, plus the two core prompts that set up safe agent context.
A prompt in MCP is a named, ready-made instruction a client can pull instead of the user or agent writing it from scratch. TAM ships 15: two core prompts that set up safe context, plus 13 packaged automation scenarios covering the everyday project-management work teams actually do.
Core prompts
`workspace_context` takes no arguments. It loads the current token's identity and workspace, then tells the agent to treat that as context, not commands, and never expand its own access beyond what's already granted. Run this before planning anything; it's the safe first move in any new session.
`triage_workflow` takes one argument, an issue id. It loads that issue (only if the connection can actually read it), then has the agent check for duplicates, assess impact and urgency, and recommend an owner, without changing anything itself. It's a read-only recommendation, not an auto-triage action.
Automation scenarios (Agent Skills)
The other 13 are ready-made scenarios for the everyday work of running a project with an AI agent:
| Scenario | What it does for you |
|---|---|
tam_orient | Gets an AI agent oriented in your workspace before it touches anything: what it can see, what it's allowed to change. |
tam_capture_work | Turns a rough idea, a meeting note, or a bug report into a properly written task with acceptance criteria. |
tam_triage_work | Reviews a new task for duplicates, checks priority, and suggests who should own it. |
tam_plan_work | Breaks a goal or a feature request down into a sequenced list of tasks. |
tam_delivery_loop | Picks up a task, does the work, and hands it back for review, following through to done. |
tam_backlog_grooming | Cleans up a messy backlog: flags stale tasks, fills in missing detail, reprioritizes. |
tam_sprint_planning | Puts together a sprint from the backlog that actually fits your team's capacity. |
tam_sprint_operations | Keeps a sprint on track day to day: flags blockers, chases anything at risk of slipping. |
tam_review_readiness | Checks whether a task is genuinely ready for review before you ask a person to look at it. |
tam_release_management | Pulls together a release checklist: what's done, what's still blocking, what's left. |
tam_github_delivery | Links a GitHub pull request or commit back to the task it closes. |
tam_project_status | Writes a project status update from what actually happened, not from memory. |
tam_manage_knowledge | Keeps wiki pages in sync with what's actually true in the project right now. |
Argument validation
Every prompt argument is checked before the prompt runs: required arguments must be present, and any string argument is capped at 2,000 characters. A prompt call with a missing required argument or an oversized one is rejected outright, not silently truncated.
Calling a prompt vs. calling a tool
A prompt doesn't change anything by itself; it returns instructions and context for the agent to act on, using the tools covered in MCP Tools. If you want TAM to actually create an issue or move a sprint, that's still a tool call. A prompt just gets the agent oriented first, with the right resources already loaded.