Agent Skill to Converge
Agent Skill to Converge is a verified SKILL blueprint on Promptally in Fullstack & Backend Engineering optimized for leading AI models. Assess the current codebase against a feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks so implementation can be completed. Post-implementation gap analysis. Adapted from github/spec-kit (speckit-converge) to run standalone, without the specify CLI. Use after a phase or feature has been implemented and you want to close the remaining gaps. Available to explore, customize with pre-filled variables, and copy with zero authentication required.
Technical Specifications
| Specification | Value |
|---|---|
| Artifact type | SKILL (Fullstack & Backend Engineering) |
| Compatible engines | Universal / Model Agnostic |
| Input variables | None (pre-filled) |
| Approximate tokens | 1,036 |
| Source file | SKILL.md |
| Access tier | Free public (MIT) |
Pre-Filled Template
---
name: converge
description: >
Assess the current codebase against a feature's spec, plan, and tasks, then append
any remaining unbuilt work as new tasks so implementation can be completed.
Post-implementation gap analysis. Adapted from github/spec-kit (speckit-converge)
to run standalone, without the specify CLI. Use after a phase or feature has been
implemented and you want to close the remaining gaps.
---
# Converge
After implementation, compare what was actually built against what the spec, plan,
and tasks said should be built, and append the remaining work to the task list. The
intent artifacts (spec, plan, tasks) are the single source of truth; the constitution
is the governing constraint.
## Hard rules
- **Append-only.** The ONLY write is appending a `## Phase N: Convergence` section to
the end of `tasks.md`. Never modify the spec, plan, or code. Never renumber or
rewrite existing tasks.
- **Not a diff tool.** Do not use `git diff`, branch compare, or VCS history. Assess
the current files on disk against the intent artifacts.
- **Require the artifacts.** If `spec.md` / `plan.md` / `tasks.md` (or equivalents)
are missing, stop and report what is missing.
- **Constitution `MUST` violations are `CRITICAL`.**
## Inputs
- `ARTIFACTS` (optional): paths to the intent artifacts. Discover defaults:
`spec.md`, `plan.md`, `tasks.md`, `.agents/plans/*/*-plan.md`.
- `CONSTITUTION` (optional): `.agents/constitution.md`; fallback `AGENTS.md`.
- `TASKS_FILE` (optional): where to append, default `tasks.md` next to the plan.
## Gap types
- `missing` — intended work with no implementation found.
- `partial` — implemented but incomplete against the acceptance criteria.
- `contradicts` — implemented differently than specified.
- `unrequested` — code exists that no artifact requested (surface for awareness,
never delete).
## Severity
- `CRITICAL` — violates a constitution `MUST`, or missing/contradicts work that
blocks the baseline of a priority-1 user story.
- `HIGH` — gap in a core requirement or success criterion.
- `MEDIUM` — secondary gap, or ambiguous `unrequested`.
- `LOW` — polish.
## Procedure
1. **Build the intent inventory.** Extract stable keys from the artifacts:
`FR-###`, `SC-###`, `AC#`, user-story IDs, plan decisions, constitution principles.
2. **Map the code scope.** Collect the files referenced by the plan and tasks; scan
them (plus keyword search) to see what exists.
3. **Assess.** For each requirement and task, classify the finding:
`missing` / `partial` / `contradicts` / `unrequested`, with a severity.
4. **Present findings first.** Show the table, then append — never the reverse:
```
## Convergence Findings
| ID | Gap Type | Severity | Source | Evidence | Remaining Work |
|----|----------|----------|--------|----------|----------------|
| C1 | missing | HIGH | FR-002 | no files | Implement login flow |
```
Include metrics (requirements assessed, tasks complete, findings by severity).
5. **Append remaining work** to the end of `tasks.md`:
```
## Phase N: Convergence
- [ ] T042 <imperative description> per <source-ref> (<gap-type>)
```
- `<source-ref>` examples: `FR-003`, `SC-002`, `US1`, `AC2`,
`plan: decision`, `Constitution II`.
- New task IDs continue from the highest existing task ID (`T042` after `T041`),
zero-padded to three digits. Constitution-derived tasks go FIRST.
- If a `## Phase N: Convergence` section already exists, create the next numbered
phase below it instead of merging.
6. **If everything is complete:** leave `tasks.md` byte-for-byte unchanged and
report `✅ Converged — the implementation satisfies the spec, plan, and tasks.`
## Output
The findings table and (when gaps exist) the appended `## Phase N: Convergence`
section. Never modify any file other than appending to `tasks.md`.
## Related skills
- analyze — cross-artifact consistency analysis, run before implementation.
- plan-phases-create — produces the plan whose gaps converge closes.
- plan-phases-implement — executes phases; converge feeds it new pending tasks.
How to use this skill
- Review the pre-filled example and source files above.
- Copy the template or download
SKILL.md, then use it in your model, IDE, or workflow tool. - Compare your result with the sample output when one is provided.
Frequently Asked Questions
How do I use Agent Skill to Converge?
Open the skill on Promptally, review the pre-filled example, then copy the template or download SKILL.md and paste it into your model, IDE or workflow tool. Agent Skill to Converge is a static blueprint with no dynamic variables. Copy it and use it as-is.
Is Agent Skill to Converge free?
Yes. Agent Skill to Converge is free to view, copy and download on Promptally without creating an account. It is published under the MIT license.
What variables does Agent Skill to Converge need?
Agent Skill to Converge is a static blueprint with no dynamic variables. Copy it and use it as-is.