ANCHOR: Uplink Issues — scope, rules, and execution contract

[lebduska]

ANCHOR: Uplink Issues — scope, rules, and execution contract

Scope

• Uplink Issues is a narrow MCP server.
• Its sole responsibility is to expose issue-tracker operations to ChatGPT.
• Supported operations are limited to:
  • read issues
  • create issues
  • update issues
  • comment on issues
  • close / reopen issues
• Anything outside this responsibility is out of scope.

---

Core guarantees

• Uplink Issues performs no autonomous actions.
• All write operations require explicit user confirmation.
• Every side effect is predictable from the tool schema.
• No operation modifies external state implicitly.

---

Prohibited behavior

Uplink Issues MUST NOT:

• act as an autonomous agent
• infer or guess user intent
• perform background automation or scheduling
• bundle multiple side effects into a single opaque action
• retry failed write operations silently
• mutate external systems without user-visible confirmation

If a feature requires guessing, it is out of scope.

---

Tools over agents

• All functionality is exposed as explicit tools.
• Tool inputs are structured and minimal.
• Tool outputs are inspectable.
• If a user cannot understand a tool’s effect by reading its schema, the design is invalid.

---

Minimalism as a constraint

• Minimal surface area is a safety property.
• Fewer tools are preferred over more tools.
• Smaller schemas are preferred over larger schemas.
• Read-only operations are preferred over write operations.
• Features that can be postponed must be postponed.

---

Source of truth

• The only sources of truth are:
  • the code in this repository
  • issues and pull requests in this repository
• There is no external specification.
• If something is not visible in the repo, it does not exist.

---

Security constraints

• No secrets are stored.
• No secrets are logged.
• Tokens must have minimal scopes.
• Failure must be explicit and loud.
• Silent fallback behavior is forbidden.

A failed operation is preferable to an ambiguous one.

---

Scope evolution

• Scope changes must be deliberate.
• Any proposal that expands scope must:
  • state which constraint it relaxes
  • justify why the relaxation is acceptable
• Unintentional scope drift is considered a bug.

---

Design axiom

Uplink Issues should feel boringly predictable.

Cleverness is a liability.