Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

Back to skills

Run Epic

ASecurity

Submit a locked ledger-native ore epic to Forged for durable frontier execution, mechanically integrate clean child slices, and stop at one draft PR to the default branch or an explicit input-required state. Use when the operator explicitly invokes /forged:run-epic.

2 stars
0 votes
0 copies
0 views
Added 9/3/2026
developmentbashgitsecurity

Works with

terminal

Security Analysis

A100/100

Scanned 9/3/2026

Install to Claude Code

$npx -y skills add tcashel/forge --skill run-epic --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Run Epic?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Run Epic
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/tcashel-run-epic/badge)](https://www.skillsdirectory.com/skills/tcashel-run-epic)

More formats (shields.io, HTML) on the badges page.

Download Zip
Files
SKILL.md
---
name: run-epic
description: "Submit a locked ledger-native ore epic to Forged for durable frontier execution, mechanically integrate clean child slices, and stop at one draft PR to the default branch or an explicit input-required state. Use when the operator explicitly invokes /forged:run-epic."
---

# /forged:run-epic

Position: locked reviewed epic -> execution handoff. Next: durable execution
under Forged's supervisor ore pass, then human adjudication of one draft PR.

Boundary: the lead session verifies the ledger plan, makes judgment calls, and
performs the explicit start/submit handoff. After submit, Forged owns frontier
scheduling, provider attempts, integration custody, child-run controllers, and
evidence. The lead remains responsible for operator conversation and final
judgment.

## Preflight — one read-only command

`forged epic preflight` rehearses every check `epic start` will enforce and
returns the identity tuple a start would freeze, creating nothing:

```bash
TARGET_REPO="$(cd "$(git rev-parse --show-toplevel)" && pwd -P)"
forged epic preflight --epic "$EPIC_ID" --repo "$TARGET_REPO" \
  --profile "${PROFILE:-standard}" --roster "${ROSTER:-default}" --rolling
```

The report's `checks` cover the repository checkout, base ref (a **bare** branch
name existing on origin, never `origin/main`), epic work-item shape, child spec
completeness versus honest planning stubs, definition and rolling package
compiles, provider binaries, and `gh` authentication. Fix every `ok: false`
check before starting; each failure names its cause.

Show the operator the returned `identities` tuple for approval before the
mutating start: normalized `baseRef`, `integrationBranch`, `assuranceStage`, and
each child's `runId`, branch, and `worktreePath`.

For judgment calls preflight cannot make, read the epic and every child in the
frozen inventory with exact-id bounded reads:

```bash
forged work show --id "$EPIC_ID"
for CHILD_ID in <every exact id from preflight's frozen child inventory>; do
  forged work show --id "$CHILD_ID"
done
```

Validate the epic and each frozen child's complete `notes`: no unchecked
question and no unresolved critique CRUX may remain. Validate every child's
reported `parent-child` and `blocks` dependency edges against the frozen
inventory, intended frontier, and ordering; a missing, extra, or contradictory
edge fails closed before approval or execution. Use the frozen child inventory
instead of recursive expansions that exceed host output budgets.

For the epic and every frozen child, critique output in `notes` must explicitly
disposition every finding, recommendation, CRUX, and open question. Each
accepted item must be folded into the normative fields; each rejected item must
retain its reason. Checkbox-free critique prose is not evidence of
adjudication. If any critique output lacks those dispositions, the epic is not
dispatchable: return the affected record to `/forged:adjudicate` before any
epic start. Keep the unchecked-checkbox gate in addition to this disposition
check.

## Freeze a proportional profile

Use standard assurance for ordinary code slices: deterministic gates, one
independent review, and a bounded remediation loop. Escalate only the children
or seams whose security, data, concurrency, or compatibility risk warrants
more. The profile's loop budget is the stopping rule.

## Typed handoff

The ordered handoff contains no spec-file argument:

```bash
forged epic start --epic "$EPIC_ID" --repo "$TARGET_REPO" \
  --profile "${PROFILE:-standard}" --roster "${ROSTER:-default}" --rolling
forged epic submit --epic "$EPIC_ID"
```

The start output must match the approved preflight identities. Submit the exact
epic id and return immediately. Submit authorizes the epic's desired row; the
supervisor ore pass reconciles it and child runs use the ordinary detached run
controllers. Do not build a second watcher loop around either lifecycle.

## Execution and terminal boundary

Forged may auto-merge only mechanically clean, accepted child slices into the
epic integration branch. It must reconcile newly ready planning stubs with
merged reality before dispatch. It must not merge the default branch.

The terminal result is exactly one draft integration-to-default PR with
evidence for human adjudication, or an explicit input-required stop naming the
failed gate, exhausted budget, stale assumption, conflict, or missing authority.

Report the epic id, frozen children, profile, integration/default branches, and
desired/pass status, then include these read-only reconnect commands:

```bash
forged overview --epic "$EPIC_ID"
forged epic status --epic "$EPIC_ID"
forged events --run "$EPIC_ID" --limit 200
```

Use a child run id returned by epic status to inspect an active provider
session, then use the attempt id returned by session inventory:

```bash
forged session list --run "$CHILD_RUN_ID"
forged session read --attempt "$ATTEMPT_ID" --lines 120
```

## Recovery

- A child `restart-budget-exhausted` condition belongs to that child run: read
  its recorded failure and use the exact run-scoped recovery verb. Resubmitting
  the epic does not reset a child run's controller budget.
- `BEADS_CONTENTION` from start, revise, or resume means the supervisor ore pass
  currently owns the epic's desired row. Back off and observe with `epic
  status`; retry only the refused control verb after that bounded pass releases
  its claim.

When `inputRequired.childId` names a child, first adjudicate that child's native
work-item fields through the lead session. Then clear only that recorded hold;
the resolution wakes the existing desired epic:

```bash
forged epic resolve --epic "$EPIC_ID" --child "$CHILD_ID" \
  --note "$RESOLUTION_NOTE"
```

For an epic-level input requirement, fix the underlying condition first, then
resolve the hold without `--child`:

```bash
forged epic resolve --epic "$EPIC_ID" --note "$RESOLUTION_NOTE"
```

## Never

- Do not synchronize an external tracker, auto-route, install software, or
  change protocol.
- Do not create an implicit default-branch approval.
- Do not replace the supervisor pass with a polling or watcher loop.

Attribution

tcasheltcashel
View sourceMore from tcashel →
SSkills DirectorySkills Directory

Your tool, in front of Claude Code builders.

3 founder slots · $299/mo · GSC-verified traffic · sponsors can never buy grades.

See placements

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Your tool, in front of Claude Code builders.

3 founder slots · $299/mo · GSC-verified traffic · sponsors can never buy grades.

See placements

Related Skills

Browser Extension Developer

Use this skill when developing or maintaining browser extension code in the `browser/` directory, including Chrome/Firefox/Edge compatibility, content scripts, background scripts, or i18n updates.

281612 votes

Seo Optimizer

SEO optimization with keyword analysis, readability assessment, technical validation, content quality. Use for search rankings, blog posts, content audits, or encountering keyword density, readability scores, meta tags, schema markup errors.

2132 votes

Google Official Seo Guide

Official Google SEO guide covering search optimization, best practices, Search Console, crawling, indexing, and improving website search visibility based on official Google documentation

1862 votes

Tanstack Start

Build a full-stack TanStack Start app on Cloudflare Workers from scratch — SSR, file-based routing, server functions, D1+Drizzle, better-auth, Tailwind v4+shadcn/ui. Use whenever the user mentions TanStack Start, asks to scaffold a full-stack Cloudflare app with SSR, wants an SSR dashboard, or asks for a React 19 + Cloudflare Workers app with file-based routing and server functions — even if they don't name TanStack Start specifically. No template repo — Claude generates every file fresh per ...

9881 votes

Pentest

PTES-aligned adversarial security audit for backend, frontend, and mobile applications. Produces a CVSS-scored Hacker Report with verified PoCs and phased remediation.

5491 votes
View all in development →