Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsBlogPro
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
  • Authors
  • 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
  • Chrome Extension
  • Skill Manager

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Agent Context

ASecurity

Use when controlling FalconDeck through its falcondeck MCP tools - creating or managing automations (scheduled agent instructions), changing agent-control settings, or inspecting runs and control history.

5 stars
0 votes
0 copies
1 views
Added 9/23/2026
developmentexpress

Works with

climcp

Security Analysis

A100/100

Pro scans all 10 files and shows the line behind each finding

Scanned 9/23/2026

$npx -y skills add jamesblackwell/falcondeck --skill agent_context --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Agent Context?

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

Security grade badge for Agent Context
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/jamesblackwell-agent-context/badge)](https://www.skillsdirectory.com/skills/jamesblackwell-agent-context)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
Files
SKILL.md
---
name: falcondeck-control
description: Use when controlling FalconDeck through its falcondeck MCP tools - creating or managing automations (scheduled agent instructions), changing agent-control settings, or inspecting runs and control history.
---

# FalconDeck control

You are running inside FalconDeck, a local-first control plane that
orchestrates coding agents (Codex, Claude, ACP CLIs) from desktop and mobile
clients. FalconDeck exposes exactly three MCP tools for inspecting and
controlling itself:

| Tool | Purpose |
| --- | --- |
| `falcondeck_search` | Discover capabilities: operation ids, schemas, examples, constraints. |
| `falcondeck_get` | Read settings, automations, run history and recent control changes. |
| `falcondeck_execute` | Execute one registered operation with validated arguments. |

Only registered operation ids can be executed; guessing internal paths does
not work. The catalogue is small — search it rather than memorising it.

## Core loop

1. `falcondeck_search` with a natural-language query (for example
   `automation create`) to find the operation id.
2. `falcondeck_search` again with `operation: "<id>"` and `detail: "full"` to
   get the complete schema and worked examples before executing.
3. `falcondeck_execute` with the operation id and validated arguments.

## Revisions

Every automation definition mutation (`automation.update`, `automation.pause`,
`automation.resume`, `automation.delete`) requires `expected_revision` — the
revision you read from a prior `falcondeck_get`. Stale revisions fail with
`revision_conflict` plus the current revision; re-read and retry.
`automation.run_now` never requires a revision.

## Idempotency

`falcondeck_execute` accepts an `idempotency_key` (8-128 characters) scoped to
origin + provider + operation. Retrying an operation with the same key and
identical arguments replays the original result instead of duplicating the
effect; differing arguments conflict. Use a fresh key (for example a UUID) for
each genuinely new operation.

## Worked example

User: "Every weekday at 8am Europe/London, use Codex in my QuizGecko
workspace to review my inbox and surface anything requiring attention."

1. `falcondeck_search` with `query: "automation create"` returns the
   `automation.create` capability with a summary schema.
2. `falcondeck_search` with `operation: "automation.create"`, `detail: "full"`
   returns the full argument schema and example payloads.
3. `falcondeck_execute` with:

```json
{
  "operation": "automation.create",
  "arguments": {
    "name": "Weekday inbox review",
    "description": "Surface inbox items requiring attention",
    "task": { "prompt": "Review my inbox and surface anything requiring attention." },
    "trigger": {
      "kind": "cron",
      "cron": "0 8 * * 1-5",
      "timezone": "Europe/London"
    },
    "target": {
      "workspace_path": "/Users/james/www/sites/quizgecko",
      "provider": "codex",
      "thread": { "kind": "managed" }
    }
  },
  "idempotency_key": "weekday-inbox-review-v1"
}
```

User (mid-conversation): "Check back on this in a couple of hours."

The user expects the follow-up in this conversation, so the target uses
`"thread": { "kind": "current" }`. FalconDeck pins it to the thread the
request came from; you do not need to know your own thread id, and you
should not pass `workspace_path` or `provider` for another conversation:

```json
{
  "operation": "automation.create",
  "arguments": {
    "name": "Check back on the campaign",
    "trigger": { "kind": "once", "run_at": "2026-08-17T15:00:00+01:00" },
    "task": {
      "kind": "prompt",
      "instruction": "Check back on the email campaign work from earlier in this thread and report status."
    },
    "target": {
      "workspace_path": "/Users/james/www/sites/quizgecko",
      "provider": "codex",
      "thread": { "kind": "current" }
    }
  },
  "idempotency_key": "campaign-check-back-v1"
}
```

The created automation survives daemon restarts and dispatches through
FalconDeck's normal threads and turns; the transcript stays in the native
agent and FalconDeck keeps only bounded run metadata.

## Automation reference

- **Triggers**: one-time (RFC 3339 with offset), five-field cron in an IANA
  timezone, or a fixed interval (minimum 60 seconds). Six-field cron
  expressions are rejected. Ambiguous local times (DST fold) run once at the
  earlier instant; skipped local times are skipped.
- **Conditional tasks**: a task with a `no_action_marker` (for example
  `FALCONDECK_NO_ACTION`) classifies a run as `succeeded_no_action` when the
  final reply is exactly that marker — ideal for "only tell me if something
  needs attention" automations.
- **Targets**: a canonical workspace path (never a runtime workspace id), an
  open provider id, and a `thread` object whose `kind` is one of:
  - `current` — continue in the thread making the request. Default for
    anything the user asks for from inside a conversation: "check back
    later", "remind me here", "re-run this tomorrow". Resolved to `existing`
    on creation; the thread's provider and workspace override the supplied
    ones.
  - `managed` — one dedicated thread FalconDeck creates and reuses. For
    standing schedules unrelated to the current conversation.
  - `existing` — a specific `thread_id` the user chose.
  - `new_each_run` — a clean thread every run. Only when the user wants no
    prior context each time.
  If the user's wording is ambiguous, prefer `current` and say which thread
  the run will land in.
- **Policies**: concurrency (`skip` default, `queue_one`, `allow`) and
  misfire (`skip` default, `run_once`). No automatic retries: provider tasks
  can have external side effects.

## Reading state

`falcondeck_get` resources: `agent_control.settings`, `automations` (list),
`automation` (one, includes instructions and revision), `automation.runs`
(history), `control.audit` (recent control changes). Automation instructions
appear only when a single automation is read explicitly.

## Notes

- Control operations have the same authority as the FalconDeck user; there is
  no separate permission system. Secret-bearing fields are redacted from
  responses.
- If a control request fails with `interface_disabled` or
  `provider_disabled`, conversational control has been switched off in
  FalconDeck's settings; say so rather than retrying.
- Use FalconDeck control when the user asks about automations, schedules, or
  FalconDeck settings. Session MCP tools such as
  `falcondeck_suggest_follow_ups` are separate: use those during ordinary
  turns without waiting to be asked.

Attribution

jamesblackwelljamesblackwell
View sourceSee grades on GitHubMore from jamesblackwell →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

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

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Clean Code

Pragmatic coding standards - concise, direct, no over-engineering, no unnecessary comments

304955 votes

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.

286712 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.

2222 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

Writing Plans

Use when you have a spec or requirements for a multi-step task, before touching code

2927051 votes
View all in development →