Agent skill

Spectre Plan

by joenandez in joenandez/spectre

Create a repository-grounded aligned draft after confirmed Scope, present its concise alignment brief, and hand it to Execute preflight.

MITAuto-check passed

Install Spectre Plan

skills CLI
$ npx skills add joenandez/spectre --skill spectre-plan -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install joenandez/spectre spectre-plan --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/joenandez/spectre.git skills-src && mkdir -p .claude/skills && cp -r skills-src/plugins/spectre/skills/spectre-plan .claude/skills/spectre-plan && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
spectre-plan
GitHub stars
164
Token cost
~1.7k tokens
SKILL.md length
742 words
Files
5 (incl. references)
Skills in repo
26
Repo updated
First seen
Licence
MIT

At a glance

Create a repository-grounded aligned draft after confirmed Scope, present its concise alignment brief, and hand it to Execute preflight.

  • Works in 8 steps: Unresolved… → Scan once; invoke… → Gather proportional evidence: XS/S local… → …
  • Execution of approved artifacts
  • SKILL.md covers Purpose, Inputs, Working Set and Outputs + DONE, plus 3 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Spectre Plan is an agent skill from joenandez/spectre. Create a repository-grounded aligned draft after confirmed Scope, present its concise alignment brief, and hand it to Execute preflight. Do not use for scoping, bug diagnosis, read-only work, or execution of approved artifacts.

Its SKILL.md is about 1.7k tokens, which your agent loads only when the skill is triggered. The skill folder holds 5 other files, including reference files (for example `references/create-plan.md`, `references/estimation-guidance.md` and `references/high-level-design-gate.md`).

The repository describes itself as: Spectre is a battle-tested 6 stage agentic engineering workflow for building complex, production-ready features. You stay in the loop for the decisions that matter. Your agents…. The licence is MIT.

When your agent uses it

  • Execution of approved artifacts

Example prompts

  • “/spectre-plan”

Workflow steps

8 steps, taken from the first numbered list in SKILL.md.

  1. Unresolved journeys/states/copy/accessibility route to spectre-ux; load-bearing interaction/layout validation routes to spectre-prototype.
  2. Scan once; invoke Skill(spectre-plan-route) in initial mode. Never print its record; its DONE and limits end at the record. In the same…
  3. Gather proportional evidence: XS/S local except probe; M ≤2 relevant @spectre:finder, @spectre:analyst, or @spectre:patterns; L/XL…
  4. Read references/minimum-solution.md after initial classification. From accepted evidence, select incumbent-first and persist ## Minimum…
  5. Invoke Skill(spectre-plan-route) in observed mode with completed selection before drafting; never print its record, and continue to step 6…
  6. Draft once by loading ${CLAUDE_PLUGIN_ROOT}/skills/spectre-plan/references/create-plan.md with the observed route-mapped depth: XS → light…
  7. Validate Routing Observations and owned concepts conform to selected record; emit plan.reclassified with plan hash, records, regret…
  8. From specs/plan.md derive requested outcome, approach, material decisions, Scope/anti-scope boundaries, credible risks, verification…

What it can do on your machine

Read from SKILL.md and the folder at commit 931c9e8. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Spectre Plan loads about 1.7k tokens when it runs, and up to ~5.6k if it reads all its reference files. Until then it costs about 60 tokens; SKILL.md has 742 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~60
When it runs · the whole SKILL.md, loaded when a task matches
~1.7k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~5.6k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from joenandez/spectre at commit 931c9e8, republished under its MIT licence (© joenandez). 742 words, ~1,675 tokens.

Download SKILL.mdSave it as .claude/skills/spectre-plan/SKILL.md (or your agent's skills folder). This skill also uses 4 other files; get the full folder from GitHub.
name
spectre-plan
description
Create a repository-grounded aligned draft after confirmed Scope, present its concise alignment brief, and hand it to Execute preflight. Do not use for scoping, bug diagnosis, read-only work, or execution of approved artifacts.
user-invocable
true
disable-model-invocation
true

plan

Purpose

Turn confirmed Scope into the smallest draft and handoff. spectre-plan-route classifies; primary owns plan. Scope remains immutable.

Inputs

  • $ARGUMENTS: confirmed Scope—thread or managed root/descendant. Repository changes without Scope return to spectre-scope; bugs use spectre-fix. Read-only diagnosis/review, release, and approved-artifact execution remain direct.
  • Existing Scope/PRD, ux.md (preferred) or legacy specs/ux.md, task_context.md, and prior plans.

Working Set

  • Reuse a managed FEATURE_ROOT only when explicit/current-thread evidence ties it to this work (physical directory wins; never branch/recency/lifecycle/scans); distinct work ignores ambient roots. Otherwise, including on collision, standalone MUST first load and follow @skill-spectre:spectre-feature-root through DONE; orchestrated calls escalate. Keep writes beneath it and pass it unchanged.
  • Immutable canonical Scope: concepts/scope.md when present, else confirmed thread; never change it without explicit scope-change approval.
  • Read references/estimation-guidance.md when applicable; each estimate excludes waits and never delays or blocks the gate.
  • Reuse current-request knowledge results/loads; otherwise follow Project knowledge routing. Search actual task across mixed knowledge/work previews; assess, then exact-load applicable records. Standalone #tag: exact search --tag, assess previews, then exact-load applicable matches; never guessed loads. Refine only for an unresolved question or new subject. Keep IDs/revisions in task_context.md; previews/unloaded candidates are not evidence.

Outputs + DONE

Every route writes specs/plan.md and task_context.md with Scope/authority-bound ## Minimum Solution Selection; no review/task/index/task-review artifacts.

DONE when task_context.md binds initial/observed routing to draft/authority hashes; Scope is unchanged; plan.completed records draft; handoff gives alignment/Trade-offs; telemetry is complete|degraded, or authority/scope stops it.

Method / guardrails

Child skill DONE is internal: resume the next Plan step in this turn. Only Plan's own handoff or escalation ends it.

  1. Unresolved journeys/states/copy/accessibility route to spectre-ux; load-bearing interaction/layout validation routes to spectre-prototype.
  2. Scan once; invoke Skill(spectre-plan-route) in initial mode. Never print its record; its DONE and limits end at the record. In the same response, run spectre-workflow plan start with root/scope hash, record, size/route, reasons, design/probe flags, boundaries; retain PLAN_RUN_ID for plan.started. Report size/rationale in one progress line beside that call; continue to step 3.
  3. Gather proportional evidence: XS/S local except probe; M ≤2 relevant @spectre:finder, @spectre:analyst, or @spectre:patterns; L/XL necessary dimensions. Live external captures belong to Execute, never Plan's evidence dispatch. Dispatch no dimension cited by path, knowledge ID/revision, or thread decision; evidence: SUFFICIENT with uncertainty below HIGH and full citation gathers locally at any size. The fresh evidence-only challenger is wave-independent: M triggers it for durable state, identity, public contract, migration, dependency, or workflow/lifecycle; L/XL always, with the wave or alone. Primary persists accepted evidence/IDs/revisions in task_context.md. @spectre:web-research only decides external API/framework. Refine knowledge search after affected files are known; load a work body only for a stated question, never reload an unchanged revision, and pass workers compact findings with provenance.
  4. Read references/minimum-solution.md after initial classification. From accepted evidence, select incumbent-first and persist ## Minimum Solution Selection in task_context.md, bound to Scope/authority/accepted evidence. XS/S decide locally; primary applies the canonical simpler-wins result.
  5. Invoke Skill(spectre-plan-route) in observed mode with completed selection before drafting; never print its record, and continue to step 6 in the same response. Plan automatically uses the observed route with no paid rerun or user tier gate unless Scope, explicit design, or authority conflicts.
  6. Draft once by loading ${CLAUDE_PLUGIN_ROOT}/skills/spectre-plan/references/create-plan.md with the observed route-mapped depth: XS → light --no-review --execution structured; S inherits it; M/L → standard --no-review --execution structured; XL → comprehensive --no-review --execution structured. Never review/task here. XS uses light, not xs, so no Execution Mode: direct.
  7. Validate Routing Observations and owned concepts conform to selected record; emit plan.reclassified with plan hash, records, regret. Append observed record to task_context.md with draft raw-byte hash plus Scope/authority hash; emit unchanged plan.completed artifact hashes/counts/planning elapsed. Telemetry failure is degraded and never blocks handoff.
  8. From specs/plan.md derive requested outcome, approach, material decisions, Scope/anti-scope boundaries, credible risks, verification intent, and ordering/shared-contract constraints. Immediately before Execute show Trade-offs verbatim (None allowed): Execute accepts them; feedback revises the draft. On Scope change, selected-design conflict, missing irreversible decision, or unavailable authority, withhold the handoff. Otherwise return exactly one copy-ready fenced command: /spectre:spectre-execute <repo-relative plan.md> --origin plan --preflight-plan <xs|light|standard|comprehensive>. observed XS → xs; S → light; M/L → standard; XL → comprehensive. Launching that command is the user's alignment signal. Never pass --orchestrated or wait for delivery insurance.
Show full SKILL.md (57 more words)Show less

Handoff

HandoffDetails
🧭 Current phaseDone
📦 What was just doneResult
▶️ Proposed next stepRender resolved action.

Execute owns assessment/review/tasks/gates/resume. Never generate a goal prompt. Render Execute: resolved absolute plan path, --origin plan, resolved preflight depth.

Escalate-If

  • Scope/root conflict; Scope/design contradiction, missing irreversible decision, unavailable authority, or invalid artifacts.
  • Telemetry/estimate failure never blocks; report degradation and continue.

© joenandez, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 4 other files (references) in plugins/spectre/skills/spectre-plan of joenandez/spectre.

  • SKILL.md
  • references/create-plan.md
  • references/estimation-guidance.md
  • references/high-level-design-gate.md
  • references/minimum-solution.md

Open the folder on GitHubat commit 931c9e8

Compare with similar skills

Spectre Plan next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Spectre Plan compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Spectre Plan this skilljoenandez/spectre164—~1.7kAutomated safety check: PassMIT
Ccs Alignthedotmack/claude-mem99k—~6.1kAutomated safety check: PassApache-2.0
Grounded CitationsNousResearch/hermes-agent252k—~3.1kAutomated safety check: PassMIT
Bio Read Alignment Bowtie2 AlignmentGPTomics/bioSkills1.2k1 repos~3.6kAutomated safety check: PassMIT
Bio Read Alignment Star AlignmentGPTomics/bioSkills1.2k1 repos~4.7kAutomated safety check: PassMIT
Bio Read Alignment Hisat2 AlignmentGPTomics/bioSkills1.2k1 repos~3.8kAutomated safety check: PassMIT

Similar skills

  • Ccs Align

    thedotmack/claude-mem

    Run the CCS Align seat's hourly breathing cycle — prove the local claude-mem worker is healthy, pull needle observations through search → timeline → getobservations, land them in a seat-owned middle…

    99k GitHub stars~6.1k tokensUpdated yesterday
    Auto-check passed
  • Grounded Citations

    NousResearch/hermes-agent

    Attaches a numbered, URL-backed citation to every outside fact in an answer or document, rejecting quotes that aren't real.

    252k GitHub stars~3.1k tokensUpdated today
    Research & ScienceAuto-check passed
  • Aligns DNA short reads to a reference with Bowtie2, choosing end-to-end (whole read must align) vs local (soft-clip read ends) mode and a sensitivity preset; the de-facto aligner for ChIP-seq…

    1.2k GitHub starsUsed in 1 repo~3.6k tokens
    Research & ScienceAuto-check passed
  • Aligns RNA-seq reads to a genome with STAR, the fast splice-aware aligner whose splice-junction database (built from a GTF at sjdbOverhang = readlength-1) and two-pass mode set junction sensitivity…

    1.2k GitHub starsUsed in 1 repo~4.7k tokens
    Research & ScienceAuto-check passed
  • Aligns RNA-seq reads to a genome with HISAT2, the splice-aware aligner whose hierarchical graph FM-index runs at roughly a quarter of STAR's memory (~7 GB for human), whose SNP/haplotype graph index…

    1.2k GitHub starsUsed in 1 repo~3.8k tokens
    Research & ScienceAuto-check passed
  • NLP Alignment

    aiming-lab/AutoResearchClaw

    Best practices for LLM alignment techniques including RLHF, DPO, and instruction tuning.

    15k GitHub stars~300 tokensUpdated 1 mo ago
    AI & LLM EngineeringAuto-check passed

More from joenandez/spectre

All 26 skills in this repo
  • Verify Spectre

    joenandez/spectre

    A skill your agent uses when validating that the Spectre plugin itself still works after ANY change to this repo — skills, prompts, agents, hooks, the CLI, the Codex mirror, or a release.

    164 GitHub stars~1.5k tokensUpdated 8 days ago
    Auto-check passed
  • A skill your agent uses when modifying the Codex SPECTRE install flow, SessionStart continuity, project skill syncing, registry injection, or Codex-specific runtime files.

    164 GitHub stars~2.2k tokensUpdated 8 days ago
    Auto-check passed
  • Spectre Handoff

    joenandez/spectre

    Save a quiet, branch-keyed session snapshot with continuity, active work, and optional task/todo state for automatic resume.

    164 GitHub stars~1.3k tokensUpdated 8 days ago
    Auto-check passed
  • Spectre Learn

    joenandez/spectre

    User-invoked front door for durable capture. An agent skill from joenandez/spectre.

    164 GitHub stars~1k tokensUpdated 8 days ago
    Auto-check passed
  • Spectre Rewrite Skill

    joenandez/spectre

    Rewrite a verbose SPECTRE skill into a compact, contract-form Spectre skill proposal — clear on WHAT, silent on HOW — grounded in the spectre-light distillation and freedom-tier budget.

    164 GitHub stars~3k tokensUpdated 8 days ago
    Auto-check passed
  • Spectre Rewrite Subagent

    joenandez/spectre

    Rewrite Spectre subagent definitions into compact, contract-driven agents with clear routing, least-privilege tools, output schemas, and an auditable rewrite trail.

    164 GitHub stars~1.8k tokensUpdated 8 days ago
    Auto-check passed

Questions about Spectre Plan

What does Spectre Plan do?

Create a repository-grounded aligned draft after confirmed Scope, present its concise alignment brief, and hand it to Execute preflight. Spectre Plan is an agent skill from joenandez/spectre. Create a repository-grounded aligned draft after confirmed Scope, present its concise alignment brief, and hand it to Execute preflight.

When should I use Spectre Plan?

Spectre Plan fits situations like: execution of approved artifacts.

How do I install Spectre Plan in Claude Code?

Run `npx skills add joenandez/spectre --skill spectre-plan -a claude-code`. Or copy the skill folder (plugins/spectre/skills/spectre-plan in joenandez/spectre) into .claude/skills/spectre-plan in your project. Claude Code loads it when a task matches its description.

How do I install Spectre Plan in Codex?

Run `npx skills add joenandez/spectre --skill spectre-plan -a codex`. Or copy the skill folder (plugins/spectre/skills/spectre-plan in joenandez/spectre) into .agents/skills/spectre-plan in your project. Codex loads it when a task matches its description.

Can I use Spectre Plan in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add joenandez/spectre --skill spectre-plan -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/spectre-plan, .gemini/skills/spectre-plan, .github/skills/spectre-plan and .opencode/skills/spectre-plan in your project.

What does Spectre Plan need to run?

SKILL.md names no scripts, command-line tools or credentials: Spectre Plan is instructions for the agent only.

Does Spectre Plan access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Spectre Plan safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Spectre Plan use?

Spectre Plan is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Spectre Plan use?

About 1.7k tokens (SKILL.md is roughly 6.7k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 3.9k tokens, read only when the agent opens those files.

What are the alternatives to Spectre Plan?

Skills that share tags, products or a category with Spectre Plan: Ccs Align (thedotmack/claude-mem, 99k stars), Grounded Citations (NousResearch/hermes-agent, 252k stars), Bio Read Alignment Bowtie2 Alignment (GPTomics/bioSkills, 1.2k stars) and Bio Read Alignment Star Alignment (GPTomics/bioSkills, 1.2k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Spectre Plan?

joenandez (a GitHub user) maintains it in joenandez/spectre, which has 164 GitHub stars. The repository holds 26 skills in this directory. The repository was last updated on October 2, 2026.

Source: joenandez/spectre on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.