implement-with-sol
Skill0 starsChanged 55 days agoMITView on GitHub
Explicitly opt into Astra-led implementation with bounded Sol workers during an authorized Heddle implement milestone. Use when the user selects this approach; ordinary implementation remains direct work by the lead.
In this repository7 other files it gives its agents.
- AGENTS.mdAGENTS.md
- architecture-proposal-reviewSkill
- implement-with-opusSkill
- new-featureSkill
- pre-implementation-analysisSkill
- root-cause-analysisSkill
- worktree-workflowSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 22
Run heddle orient and follow its next actions.
- Very specific: line 28
If Heddle-owned guidance is absent from the host checkout, use heddle doctor --json to locate the installation.
- Very specific: line 48
Use sol-implementer, defined in .codex/agents/sol-implementer.toml.
- Distinctive: line 79
When the user selects standard implementation, finish or stop the active worker's handoff before the lead resumes writing.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Implement with Sol
Deliver the authorized milestone as a coherent, correct, and maintainable whole. Astra leads and delegates implementation to Sol where a bounded contract and adequate evidence justify the handoff. Direct implementation remains useful when the work needs shared, changing context or the handoff adds little value.
Selection and authority
Invoke $implement-with-sol in an Astra session. This skill does not change the
lead's model or reasoning effort. The selection applies to the current milestone
unless the user specifies a different scope within the implement stage.
Run heddle orient and follow its next actions. Read
heddle/resources/implement.briefing.md and the governing records. The user's
request and governing Heddle and repository authority take precedence over this
skill and its assignments. This skill changes execution style, not scope, stage
authority, test authority, or review policy.
If Heddle-owned guidance is absent from the host checkout, use
heddle doctor --json to locate the installation. Read an editable checkout's
resources where available; otherwise use native help and routed briefings,
identifying any remaining guidance gap.
Assigning work
Choose between direct work and delegation by uncertainty, consequence, coupling, and verifiability. Retain consequential shared reasoning and integration with the lead. Give Sol a cohesive outcome with settled inputs, clear ownership, and observable completion criteria.
Build the assignment with
heddle/resources/work-package-brief.scaffold.md. Use the existing
specification, milestone scope, scaffolded tests, consumer contracts, and
applicable verification authority. Supply the context and operative
instructions the worker needs without prescribing reversible implementation
choices. A defect in those inputs is a discovery to resolve through the
governing workflow, not permission to weaken the contract.
Use sol-implementer, defined in .codex/agents/sol-implementer.toml. Its
reasoning effort is pinned to xhigh. Check the active harness's model,
context, and permission controls. Surface an unavailable agent, model, or effort
level rather than silently substituting one or claiming the requested
configuration ran. The custom agent's instructions are behavioral constraints;
workers inherit the active harness permissions.
Execution and handoff
The lead may supervise multiple Sol workers in parallel when their work packages are independent, shared contracts are settled, write ownership does not overlap, and the integration order is clear. One owner may write a given path at a time. The lead may implement a separate package concurrently under the same rule. If an overlap, dependency, or shared-contract change emerges, pause the affected package and re-brief it from current source before writing continues.
The lead tracks package ownership, dependencies, and source identities while retaining commits, Heddle mutations, governing-document changes, integration, and formal review within its existing authority. Integrate in dependency order and rerun affected evidence on the integrated source; a worker's passing result can become stale after another package lands.
Inspect the returned changes and evidence against the governing requirements and actual consumers. Resolve consequential gaps before dependent work. Reuse worker context when it remains useful; change the assignment or implement directly when that better addresses the source of uncertainty. Follow native verification, review, remediation, and completion requirements. Do not add an extra verifier without a distinct evidence need; supplementary assessment does not replace formal review.
When the user selects standard implementation, finish or stop the active worker's handoff before the lead resumes writing. Preserve its changes, inspect the resulting source, and record remaining work in the existing handoff. Changing execution style neither discards evidence nor establishes that it still applies.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

