Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/layer1labs/specsmith/agent-corenpx skills add layer1labs/specsmith --skill agent-coregit clone --depth 1 https://github.com/layer1labs/specsmithWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.00624 |
| Opus 5 | $0.00000 | $0.00312 |
| Sonnet 5 | $0.00000 | $0.00125 |
| Haiku 4.5 | $0.00000 | $0.00062 |
Grade A, and why
agent-core scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 85 lines — stays where its author put it; the contents beside it link to each section on GitHub.
SKILL.md
Agent Core Profile
Requirements Covered
Governance Requirements
- REQ-012: Least-Privilege Agent Permissions
- REQ-217: Agent Permissions Check
- REQ-220: Policy Guardrails
- REQ-244: Context Window Sizing
- REQ-245: Live Context Fill Indicator
- REQ-246: Auto Context Compression
- REQ-247: Hard Context Ceiling
Core Principle
The agent-core profile defines the fundamental capabilities and governance rules that all Specsmith agents must follow. This profile ensures that agents operate within the bounds of the Specsmith governance framework while maintaining the flexibility to adapt to different project contexts.
Profile Characteristics
Minimal Required Skills
- preflight-gate
- governed-agent-loop
- requirement-author
- testcase-author
- traceability-auditor
- context-pack-compiler
- token-budget-auditor
- skill-composer
Governance Compliance
All agents using this profile must:
- Run preflight checks before any action that could have side effects
- Follow the governed-agent-loop execution pattern
- Maintain traceability between requirements, tests, and code changes
- Compose skills dynamically based on context
- Track token usage and budget efficiency
Implementation Details
Core Loop Execution
- Detect Specsmith project context
- Run session bootstrap procedures
- Emit governance anchor
- Classify user intent (read-only, change, release, destructive, research, planning)
- For non-read-only actions, run preflight
- Load only required context and skills
- Execute smallest scoped action
- Run verification gates
- Record evidence
- Save through Specsmith
Context Management
- Build minimal context packs from project state, requirements, work items, ESDB records, and relevant skills
- Never inject the whole repository when a scoped context is sufficient
- Prefer requirement IDs, test IDs, recent WIs, and relevant files
- Exclude stale, low-confidence, tombstoned, or contradicted ESDB records
- Include stop conditions and known hazards
- Track token budget and context utilization
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 85 lines · 0 tokens per session scan A cfa2e0c9607b
agent-core is a skill published in the GitHub repository layer1labs/specsmith (7 stars, last pushed 29d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 624 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
arch-audit
Audit Claude Code architecture files against Anthropic docs and release notes, and verify internal ecosystem consistency. Run on demand to maintain compliance, catch new features, and keep the context system clean.
skill-review
Orchestrate the skill-review framework v1.2 pipeline on a target skill or cross-tier family. Runs Phase 1 preflight, Phase 2 structural review with interactive walkthrough, Phase 3 fix + rollback, optional Phase 4 external LLM review, Phase 5 integration, Phase 6 closeout. Supports full, preflight-only, and…
external-review
Run a fresh-context review of Tierward before a release. --mode=quick spawns a Claude general-purpose subagent with no project memory and an auto-bundled snapshot. --mode=full fans out the same bundle to GPT-4.1, Gemini 2.5 Pro, Mistral Large, and Perplexity Sonar Pro for cross-LLM coverage. Maintainer-only; not…
ifixai
Guide the user through an independent iFixAi audit of their own agent, checking whether it does the job it is supposed to do given their business rules and org structure. Prefer pointing it at the user's REAL deployed agent over its HTTP endpoint (its actual tools, retrieval, and governance) with --provider http…
improve-harness
Run one explicitly authorized, evidence-backed improvement to a repository's agent guidance, tools, runbooks, or validation. Use only when the user invokes $improve-harness or explicitly asks to improve the Harness after observed reusable agent friction. Do not use for ordinary product changes, speculative cleanup…
karpathy-principles
Behavioral guidelines to reduce common LLM coding mistakes — Think Before Coding, Simplicity First, Surgical Changes, Goal-Driven Execution.