AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Envelope

skill-sandsower-beislid-envelope · by sandsower

Author, approve, and export execution envelopes for AFK slices in a standalone strong-model session. Use only when the user explicitly invokes /envelope with a ticket id, approved Work Contract/spec file, or batch. Other skills may recommend it but must never auto-route into it. Ends with an approved, validated, repo-committed bundle under .beislid/exports/ that an external runner executes later.

No reviews yet
0 installs
33 views
0.0% view→install

Install

$ agentstack add skill-sandsower-beislid-envelope

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No 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.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-sandsower-beislid-envelope)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
2mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

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 →
Are you the author of Envelope? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Envelope

Turn approved planning context into exported execution envelopes: one execution-envelope-v0 per AFK-ready slice, packaged as an approved-slice-plan-export-v0 bundle that rondo run-once --manifest can execute in a fresh session with zero additional context.

Trigger is explicit only. The human chooses which model/session pays for authoring. When kickoff classifies work as multi_slice/AFK-suitable it may recommend running /envelope in a strong-model session; it never auto-routes here, and this skill never auto-routes back.

Don't use this for: interactive implementation (use kickoff), PR handoff (ready-for-review), or running exported work (that's rondo's job — Beislið owns envelope/export semantics; rondo owns execution and run evidence; no runtime service).

Project config: /.beislid/workflow.md. Probe lazily per probe-semantics.md; policy-check side effects per action-policy-protocol.md. Output copy lives in envelope-templates.md. AFK eligibility is judged against a versioned rubric, resolved repo-override-first: a beislid:envelope block in workflow.md may set rubric_path (a repo-relative .md, no ..); otherwise the skill default afk-rubric.md applies.

Read workflow.md

Read /.beislid/workflow.md and validate line 1 is exactly `; hard-fail and stop if missing or wrong (run /setup for fresh projects). Initialize the probe cache exactly as kickoff does (repohash, workflowhash, ${BEISLIDSTATEDIR:-$HOME/.local/state/beislid}/probes/.json, fresh/stale/cold). Print orientation prose from envelope-templates.md` (≤240 chars).

For durable evidence, best-effort beislid run-ledger init --skill envelope --flow envelope with ticket/branch context when known. Warn on ledger failure; never block.

Step protocol loading

Complete steps in order. At each step entry, read the step aux file and follow it as the authoritative protocol. If a step file cannot be read, hard-fail and stop:

> 🛑 Could not read skills/envelope/.md. Envelope cannot safely execute this step from memory; reinstall Beislið or restore the file.

Checklist and required outputs

  1. Intake — read step-1-intake.md. Outputs: input kind, planning context (approved structure/Work Contract or planning-skill results), candidate slices, bundle-id.
  2. Author — read step-2-author.md. Outputs: one draft execution-envelope-v0 per AFK-candidate slice with autonomy lists, proof requirements, pause conditions, dependencies, expected delivery, self-contained prompt, repo pin.
  3. Approve — read step-3-approve.md. Outputs: per-envelope verdicts (approve / reject / demote-to-HITL), approval metadata, or the zero-AFK terminal state.
  4. Export — read step-4-export.md. Outputs: validated bundle path, envelope_exported checkpoint status, commit status, handoff guidance.
  5. Reviserevision mode only, loaded straight from Step 1 in place of Steps 2–4: read step-5-revise.md. Outputs: v+1 bundle rewritten in place with supersedes: , per-envelope delta summary, delta-only re-approval verdicts (unchanged approved envelopes carry their original approval metadata forward), validated re-export.

Global tripwires

  • Fail-closed: unapproved envelopes are never exported. Zero approved envelopes means no bundle, no checkpoint, no commit — print the verdict summary and recommend kickoff for HITL slices.
  • No implementation. This skill plans, packages, and exports; it never starts the work it is packaging.
  • Manifest input means revision mode. Step 1 detects it (a JSON file whose kind/schema matches export/manifest contracts) and looks for feedback: a delivery artifact with pause_reason/review_feedback, or a non-approved bundle status. With feedback, route through step-5-revise.md; an approved manifest with no feedback is refused with the nothing-to-revise message. Never overwrite or re-author silently.
  • Existing bundle dir is a hard stop in non-revision runs. .beislid/exports// already existing means a prior export; overwriting would corrupt the supersede chain. Stop with the collision message; the user must pick a new bundle-id or delete deliberately. Revision mode is the one exception: it rewrites the same dir in place with a bumped version and supersedes hash.
  • Validator is the gate. beislid export validate must exit 0 before checkpoint or commit. Never hand-wave a failed validation.
  • Reuse planning skills (spec, break-spec, blueprint) for thinking work; never reimplement their protocols here.
  • Policy-check every side effect (workspace writes, checkpoint, git.commit); ask requires approval, deny stops with manual fallback.

Run end

Write back probed capability entries to the probe cache as kickoff does (exclude session_skip, preserve doctor-owned fields). Finalize the run ledger with verdicts, bundle path, validator evidence, and commit status when active.

Key principles

  • One strong-model session produces a self-contained bundle; a cheaper session executes it later.
  • Explicit human verdicts per envelope, in one sitting; rejects never block the batch.
  • Everything after approval is mechanical: documented JSON conventions + a read-only validator.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.