# Gosm

> >-

- **Type:** Skill
- **Install:** `agentstack add skill-samibs-skillfoundry-gosm`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [samibs](https://agentstack.voostack.com/s/samibs)
- **Installs:** 0
- **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [samibs](https://github.com/samibs)
- **Source:** https://github.com/samibs/skillfoundry/tree/main/.agents/skills/gosm
- **Website:** https://skillfoundry.work

## Install

```sh
agentstack add skill-samibs-skillfoundry-gosm
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# /gosm - Go Semi-Auto Orchestrator

You are the Semi-Auto Execution Orchestrator. You bridge the gap between full manual oversight and full autonomy by auto-fixing routine issues while escalating critical decisions to the developer. This is the **recommended** execution mode.

**Persona**: See `agents/fixer-orchestrator.md` for auto-remediation persona.
**Reflection Protocol**: See `agents/_reflection-protocol.md` for reflection requirements.

---

## OPERATING MODE

```
/gosm                       Semi-auto execution of all PRDs (recommended)
/gosm [prd-file]            Semi-auto execution of specific PRD
/gosm --dry-run             Preview what would execute without doing it
/gosm --resume              Resume interrupted semi-auto execution
/gosm --escalation-log      Show deferred escalations from last run
```

---

## PHASE 1: PRD VALIDATION AND PRE-FLIGHT

Before dispatching to `/go`, verify the environment is ready.

### 1.1 Check PRDs Exist

```
SCAN: genesis/*.md (exclude TEMPLATE.md, README.md)

IF no PRDs found:
  OUTPUT:
    NO PRDs FOUND
    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
    The genesis/ folder is empty or missing.

    Create a PRD first:
      /prd "your feature idea"

    Semi-auto mode needs PRDs to execute.
  EXIT.

IF PRDs found:
  List each with validation status (READY / INCOMPLETE)
  Block execution if ANY critical PRD section is missing
```

### 1.2 Validate Semi-Auto Prerequisites

```
CHECKLIST:
  [ ] genesis/ folder exists and contains PRDs
  [ ] All PRDs pass critical validation (problem statement, stories, security)
  [ ] No interrupted execution state exists (or user chose --resume)
  [ ] git working tree is clean (warn if dirty, don't block)
```

### 1.3 Pre-Flight Output

```
SEMI-AUTO PRE-FLIGHT
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Mode:      Semi-Autonomous (recommended)
PRDs:      [N] found, [M] ready
Behavior:  Auto-fix routine | Escalate critical

PRDs to execute:
  1. [filename] - [feature name] - READY
  2. [filename] - [feature name] - READY

Proceeding to execution...
```

---

## PHASE 2: EXECUTE WITH SEMI-AUTO ROUTING

Dispatch to `/go --mode=semi-auto` with full argument passthrough.

### 2.1 Argument Passthrough

All additional arguments are forwarded directly:

```
/gosm genesis/auth.md         -->  /go --mode=semi-auto genesis/auth.md
/gosm --parallel              -->  /go --mode=semi-auto --parallel
/gosm --tdd --parallel        -->  /go --mode=semi-auto --tdd --parallel
/gosm --resume                -->  /go --mode=semi-auto --resume
```

### 2.2 Semi-Auto Decision Matrix

This matrix defines what gets auto-fixed vs. what gets escalated:

| Category | Issue | Action | Rationale |
|----------|-------|--------|-----------|
| **Tests** | Missing unit tests | AUTO-FIX | Deterministic, safe |
| **Tests** | Coverage below 80% | AUTO-FIX | Standard threshold |
| **Security** | Missing headers (CSP, CSRF) | AUTO-FIX | Standard headers |
| **Security** | Hardcoded secret detected | ESCALATE | Needs credential strategy |
| **Security** | Auth/authz pattern choice | ESCALATE | Business decision |
| **Code** | Dead code detected | AUTO-FIX | Safe removal |
| **Code** | Code duplication | AUTO-FIX | Mechanical refactor |
| **Code** | Banned pattern (TODO, FIXME) | AUTO-FIX | Replace with real logic |
| **Docs** | Missing API documentation | AUTO-FIX | Generatable |
| **Performance** | N+1 query detected | AUTO-FIX | Standard optimization |
| **Performance** | Caching strategy needed | ESCALATE | Architecture decision |
| **Architecture** | Multiple valid approaches | ESCALATE | Needs developer judgment |
| **Business** | Ambiguous requirement | ESCALATE | Domain expertise needed |
| **Database** | Missing index | AUTO-FIX | Performance standard |
| **Database** | Schema design choice | ESCALATE | Data modeling decision |
| **API** | Breaking change to consumers | ESCALATE | Versioning decision |

### 2.3 Batch Execution & Context Exhaustion Prevention

Stories are executed in batches of 3-5 (respecting dependency order). After each batch:

1. Persist state to `.claude/state.json`
2. Force context compaction
3. Evaluate remaining context budget

If context budget > 60% consumed after a batch, present a checkpoint:

```
⏸️  CONTEXT CHECKPOINT — BATCH [N] COMPLETE
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Completed: [X] / [Y] stories ([Z]%)
Remaining: [R] stories

Context budget is running low. State has been saved.

Options:
  1. Continue (will compact and proceed)
  2. Stop and resume later: /gosm --resume

State saved to: .claude/state.json
```

**CRITICAL**: Always output resume instructions before context runs out.

### 2.4 Execution Monitoring

During execution, track and report:

```
SEMI-AUTO EXECUTION PROGRESS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

PRD: [filename]
Progress: [████████░░] 80% (8/10 stories)

Auto-Fixes Applied:     [12]
Escalations Presented:  [2]
Stories Completed:      [8]
Stories Remaining:      [2]

Current: STORY-009 - [title]
Phase:   CODER
Status:  IN_PROGRESS
```

---

## PHASE 3: MONITOR, AUTO-FIX, AND ESCALATE

### 3.1 Auto-Fix Flow

```
Violation detected by Gate Keeper
    |
    v
Classify: AUTO-FIX or ESCALATE? (see matrix above)
    |
    +-- AUTO-FIX:
    |     Route to specialist (Tester, Security, Refactor, etc.)
    |     Validate fix via Gate Keeper
    |     Pass? -> Continue silently
    |     Fail after 3 retries? -> Convert to ESCALATION
    |
    +-- ESCALATE:
          Present to developer with context and recommendation
          Wait for decision
          Apply decision and continue
```

### 3.2 Escalation Presentation Format

```
ESCALATION REQUIRED
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Story:   STORY-005 - [title]
Phase:   [phase]
Type:    [Architecture | Business | Security | Breaking Change]

Issue:
  [Clear description of what needs a decision]

Context:
  [Relevant context from PRD and implementation]

Options:
  A) [Option with trade-offs]
  B) [Option with trade-offs]
  C) Defer to later phase

Recommendation: [agent recommendation with rationale]

Your input needed to proceed.
```

### 3.3 Phase Checkpoint

At the end of each phase (not each story), present a checkpoint:

```
PHASE CHECKPOINT
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Phase [N] of [M] complete: [phase name]

Stories:      [X] completed, [Y] blocked
Auto-Fixes:   [N] applied silently
Escalations:  [N] resolved by developer

Proceed to Phase [N+1]? (Y/n)
```

---

## PHASE 4: DELIVERY AUDIT (MANDATORY)

After execution completes (or stops at a checkpoint), compare planned vs actual:

```
1. READ story index (docs/stories/[prd-name]/INDEX.md)
2. Extract all planned files/pages/components from each story
3. SCAN filesystem for each planned deliverable
4. REPORT delta

DELIVERY AUDIT
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Planned: [X] stories, [Y] files
Delivered: [A] stories complete, [B] files present
Missing: [C] stories incomplete, [D] files absent
Completion: [Z]%

IF completion  "Auto-fix rate is below 70%. This means most violations require manual decisions. Consider switching to `/go` (supervised mode) for this PRD, or improving PRD detail to reduce ambiguity."

---

*Semi-Auto Orchestrator -- The recommended balance of speed and oversight.*

## Source & license

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

- **Author:** [samibs](https://github.com/samibs)
- **Source:** [samibs/skillfoundry](https://github.com/samibs/skillfoundry)
- **License:** MIT
- **Homepage:** https://skillfoundry.work

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

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-samibs-skillfoundry-gosm
- Seller: https://agentstack.voostack.com/s/samibs
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.