Skip to content

feat(tech-writing): add client-facing-doc and bid-delivery-summary skills#17

Merged
MrBogomips merged 1 commit into
mainfrom
feature/tech-writing-doc-skills
Jun 23, 2026
Merged

feat(tech-writing): add client-facing-doc and bid-delivery-summary skills#17
MrBogomips merged 1 commit into
mainfrom
feature/tech-writing-doc-skills

Conversation

@MrBogomips

Copy link
Copy Markdown
Owner

Summary

Adds the internal/external document-transformation pair to the tech-writing plugin, which was previously an empty placeholder. The two skills are mirror images of each other, both operating on the same kind of source material (project assessments, estimations, technical analyses):

  • client-facing-doc — converts internal/reserved technical documentation into a polished client-facing deliverable. Silently removes confidential content (costs, rates, effort estimates, internal planning, AI artifacts, drafts, internal notes) while preserving and improving the customer-relevant technical substance. Produces the deliverable in docs/outbox/ plus a separate redaction audit in .aidocs/ (gitignored) so a leak can be caught before sharing.
  • bid-delivery-summary — distills the same kind of source into a concise internal commercial & delivery summary (INTERNAL USE ONLY) for sales, bid, and delivery stakeholders: scope, resources, effort, risks, dependencies, a 4-way bid-readiness verdict, and a delivery-readiness assessment. Effort-only in PD/MD by default; cost figures appear only when an approved cost model is present and explicitly authorized.

Both skills:

  • run an audience-aware section-plan approval step before generation — they propose which sections to include and which to remove based on the audience (client vs internal) and ask the user to confirm/adjust;
  • are bilingual (English/Italian) via language packs;
  • accept PDF/DOCX/PPTX/XLSX inputs via the markitdown MCP tool and offer optional DOCX output via document-skills/links-gc-xdoc;
  • follow house conventions (rich bilingual description, numbered pipeline, progressive-disclosure table, docs/outbox/ output, self-check rules).

Plugin / marketplace changes

  • tech-writing bumped 0.1.00.2.0 (both plugin.json and marketplace.json, kept in sync).
  • New tech-writing/CONNECTORS.md declaring ~~knowledge base.
  • Broadened plugin description and tags; updated README.

Test plan

  • bash tests/ci/run-structural-tests.sh — all suites pass (frontmatter valid, references resolve, connector matched, versions in sync).
  • claude plugin validate ./tech-writing --strict — passes.
  • Functional smoke test (manual, recommended in a fresh session): run each skill on a fictional source with planted markers and confirm — for client-facing-doc, no confidential content leaks into the deliverable and the audit lists every removal; for bid-delivery-summary, the INTERNAL USE ONLY header is present, the cost gate asks for confirmation when a rate card is supplied, and the readiness verdicts come from the enumerated sets.

…ills

Add the internal/external document-transformation pair to the tech-writing
plugin (previously an empty placeholder):

- client-facing-doc: convert internal/reserved technical documentation into a
  client-facing deliverable, removing confidential content (costs, rates, effort,
  internal planning, AI artifacts, drafts, notes) and producing a separate
  redaction audit in .aidocs/ for verification.
- bid-delivery-summary: distill a project assessment/estimation into an internal
  commercial & delivery summary (INTERNAL USE ONLY) with bid-readiness and
  delivery-readiness verdicts; effort-only in PD/MD unless an approved cost model
  is present and explicitly authorized.

Both skills run an audience-aware section-plan approval step before generation
(include/remove sections by audience) and support English/Italian via language
packs.

Bump tech-writing to 0.2.0; add CONNECTORS.md (~~knowledge base); update README
and marketplace.json. Structural validation (tests/ci/run-structural-tests.sh)
and `claude plugin validate ./tech-writing --strict` both pass.
@MrBogomips MrBogomips merged commit aa412c8 into main Jun 23, 2026
1 check passed

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1952f3420c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment on lines +156 to +157
Write the redaction audit to `.aidocs/<doc-name>-redaction-audit.md` using
`references/audit-template.md`, populated from the audit buffer.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Ensure redaction audits are ignored before writing

When this skill runs in a customer project that does not already have .aidocs/ in .gitignore, the redaction audit written here is a normal repo file even though the audit buffer stores snippets of removed costs/internal notes. The .gitignore in this plugin repo does not protect consumer worktrees, so the skill should verify/add an ignore rule or write the audit outside tracked paths before dumping confidential snippets.

Useful? React with 👍 / 👎.

Comment on lines +142 to +145
- Internal-status terms: `TODO`, `FIXME`, `WIP`, `draft`/`bozza`, `internal`/`interno`/`riservato`,
`confidential`/`riservato`
- AI/process leakage: prompt fragments, "as an AI", chain-of-thought phrasing ("let me think",
"step 1: I will"), references to AI tools or generated artifacts

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Narrow residual markers to process leakage

With customer-facing material that legitimately mentions a customer's internal systems (e.g., internal APIs/networks) or an AI component, this final pass removes those words/references solely because they match internal/interno or “AI tools”. That contradicts the preservation goal for architecture, dependencies, and technical substance and can silently delete valid client content; the scan should target explicit internal-only/process-leakage phrases instead.

Useful? React with 👍 / 👎.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant