parallel-agents

A workflow for dividing several independent coding jobs among multiple sub-agents, then combining their results. A sub-agent is a separate helper focused on one assigned task.

In plain words
What is it for?
It is intended for repeated work across separate engine projects, modules, pattern files, or language files, when the jobs do not depend on one another.
Why use it?
It reduces coordination time when the same kind of change must be made across unrelated files or modules.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/bablsoft/accessflow/parallel-agents
Any agent
npx skills add bablsoft/accessflow --skill parallel-agents
Clone the repo
git clone --depth 1 https://github.com/bablsoft/accessflow

Made for: Claude Code, Codex.

Per session 99 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,242 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.00099 $0.01242
Opus 5 $0.00049 $0.00621
Sonnet 5 $0.00020 $0.00248
Haiku 4.5 $0.00010 $0.00124

Measured yesterday against content hash 775201feeb22, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

parallel-agents 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 yesterday.

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.

.claude/skills/parallel-agents/SKILL.md · 103 lines

How it starts

The opening of the file, as written. The whole thing — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.

parallel-agents

AccessFlow has an unusual amount of genuinely parallel work, because the ten engine plugins are standalone projects that repeat the same shapes. This skill is for that — and only that.

When it fits

  • The nine *RowSecurityApplier classes. The canonical job. Disjoint files, identical task shape, and two of them (Cassandra, Neo4j) have a default branch so the compiler will not find them for you → patterns/engine-fanout.md.
  • Per-engine changes that follow a core.api SPI change — one agent per engines/<id>/.
  • Writing several pattern or doc files in one pass.
  • Adding the same i18n key to six locale files (though a script is usually better).
  • Per-module refactors where each module's internal/ is touched independently.

When it does not

  • Anything touching a shared file — see the conflict list below. Two agents editing types/api.ts will clobber each other, and neither will notice.
  • Work with a dependency chain. If B needs A's output, that is sequential.
  • Fewer than three jobs. The coordination overhead exceeds the saving.
  • Anything that needs a shared decision. Decide first, then fan out the mechanical part.

Shared files — never edit these from more than one agent

File Why
backend/src/main/resources/db/migration/ Version collision — two agents both pick V130. Assign the number yourself, up front.
backend/src/main/resources/i18n/messages.properties + the 6 locales Every agent appends; last write wins
frontend/src/types/api.ts One big union/interface file
frontend/src/utils/enumLabels.ts Same
frontend/src/locales/*.json Same
backend/pom.xml, frontend/package.json Dependency edits serialize
website/sitemap.xml One <url> list
.github/workflows/ci.yml One matrix
CLAUDE.md, .claude/patterns/README.md Index files

The pattern that works: fan out the per-file work, then do the shared-file edits yourself in the consolidation step. Tell each agent explicitly that the shared files are off-limits, and have it report what it would have added instead.

Read the full file on GitHub · 103 lines

Changes

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.

  1. yesterday First seen · 103 lines · 99 tokens per session scan A 775201feeb22

Subscribe to this mod's changes

parallel-agents is a skill published in the GitHub repository bablsoft/accessflow (4 stars, last pushed 2d ago), licensed Apache-2.0. It adds 99 tokens to every session and 1,242 once invoked, about $0.0005 per session on Opus 5. 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.

Related

Other skills, from other repositories

wire-capability-probe

Establish what a provider wire can and cannot carry, and fix the gateway when the answer is "we send the wrong form". Use when an upstream refuses a request (4xx/422 naming a content type, a field, a modality, a URL form), when a catalog capability is about to be written from a refusal, or when a codec translation is…

AlphaBitCore/nexus-gateway · 150 tokens

datadog-inspector-expert

Interpret datadog-inspector findings and translate Datadog monitoring, audit, log-retention, SSO, and RBAC results into GRC evidence and remediation.

GRCEngClub/claude-grc-engineering · 42 tokens

splunk-inspector-expert

Interpret splunk-inspector findings and translate Splunk retention, RBAC, audit, search ACL, and auth posture into compliance evidence and remediation.

GRCEngClub/claude-grc-engineering · 36 tokens

cross-domain-translator

Translates adjacent-domain patterns into GRC learning frames. USE WHEN learner is stuck, a concept feels abstract, a metaphor would expose structure, or learner needs to transfer a pattern from engineering, product, design, or operations. NOT FOR direct concept teaching when plain explanation is enough; use…

grcengineering/companion · 66 tokens

linmas-threat-research-analyst

Threat research skill for IOC analysis, adversary tracking, campaign reporting, and intelligence-to-detection translation.

TanKimGwan/linmas · 30 tokens

connect_polaris_catalog

Guides the agent to ask the user for their preferred authentication mode and credentials when they request to connect to the Polaris catalog, rather than using default credentials.

sankeerthnagapuri/apache-polaris-iceberg-ai-mcp · 37 tokens