Install
$ agentstack add skill-minwoo19930301-relay10-relay10-spec ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
DisciplinedRun Spec
Design the smallest change that can prove the requested outcome. Resolve facts from evidence, and ask the user only for decisions that evidence cannot make.
Clarify only what matters
- Inspect the repository before asking about filenames, commands, existing
behavior, or other facts that local evidence can answer.
- Ask only when the answer would materially change the outcome, authority,
risk, acceptance criteria, or irreversible action.
- Ask one question at a time and default to at most three questions. If a
blocking decision remains after three, stop with an unresolved contract instead of guessing or starting implementation.
- Record each material premise as
user_decision,repo_fact,
safe_assumption, or blocker, with its evidence or owner.
- Do not treat "알아서 해", "use your judgment", silence, or a broad goal as
authorization for destructive changes, publication, spending, credentials, or production operations.
Skip the interview when the request is already specific, reversible, and verifiable. State a safe assumption and continue when the assumption is inside the user's authority and cannot materially change the result.
Build the specification
- Inspect the existing behavior, conventions, tests, and user-owned changes.
- State the user outcome in one sentence.
- List non-goals and constraints, including provider, app, platform, and licensing boundaries.
- Describe two alternatives only when the choice materially affects scope or risk.
- Select one approach and explain the rejected alternative briefly.
- Define observable acceptance criteria before listing implementation files.
- Add a verification plan, rollback path, and unresolved questions.
Before handing work to relay10-build, restate the result as a Confirmed Task Contract using [the contract fields](./references/task-contract.md). A clear original request can confirm revision 1 without another ceremonial question. When clarification changed or selected the scope, ask the user to confirm the restatement. A contract with a blocker or unresolved material decision cannot authorize implementation.
After confirmation, do not silently rewrite intent. A material requirement change creates a new revision that names the prior revision and repeats only the confirmation needed for the changed decision.
Risk routing
Assess each dimension from low to high:
- complexity;
- blast radius;
- security or data sensitivity;
- verification difficulty;
- irreversibility.
Use stronger planning and review only where these dimensions justify it. A small, reversible, well-tested edit should stay small.
Output contract
Produce:
- contract status, revision, and decision provenance;
- outcome and user-visible behavior;
- current-state evidence;
- non-goals and constraints;
- allowed mutations and forbidden actions;
- chosen design and key interfaces;
- file-level implementation map;
- acceptance tests;
- risks, rollback, and follow-ups.
For plan-only requests, stop after the specification. Do not implement or publish.
This is a lightweight confirmation contract, not an immutable runtime Seed, numeric ambiguity score, ontology, resume engine, or autonomous evolution loop.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: minwoo19930301
- Source: minwoo19930301/relay10
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.