agentleFS
gootfgootf/software-agent-sop/synthesize-project-context/SKILL.md

synthesize-project-context

Skill2 starsChanged 2 months agoMITView on GitHub

Synthesize upstream artifacts (requirements, source evidence, architecture/dependency/design findings, readiness) into candidate entries for PROJECT-CONTEXT.md — the S10 synthesis step that the project-context carrier format lacks. Use when multiple upstream analyses have completed and the unified project context must be written or updated (S10 synthesize-project-context slot), or when asked to "synthesize project context" / "update project context from findings" / "write the project context from these reports".

  • Reads credentials

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 1 is distinctive, and 4 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.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read every listed artifact fully.

    line 34
  2. Very specific:

    Read existing ## Active entries of PROJECT-CONTEXT.md.

    line 49
  3. Very specific:

    Ask once: approve all / adjust (specify rows) / reject (specify rows).

    line 64
  4. Very specific:

    Write approved entries into ## Active of PROJECT-CONTEXT.md (create the file with the standard header if absent).

    line 68
  5. Distinctive:

    Before declaring done, all must hold:

    line 72

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Synthesize Project Context

When to Use

  • The S10 slot: after S03 (requirements), S05 (source evidence), S06 (architecture findings), S07 (dependency findings), S08 (design findings), S09 (readiness decision) outputs exist, before S11 planning.
  • Any bulk update of PROJECT-CONTEXT.md from multiple findings (not single-entry adds — those go through project-context add).

Inputs

  • Paths to upstream artifacts (provided at invocation): requirements / source-evidence / architecture-findings / dependency-findings / design-findings / readiness-decision. Any subset is valid.
  • Existing PROJECT-CONTEXT.md at repo root, if present.

Outputs

  • Updated PROJECT-CONTEXT.md (approved candidates appended to ## Active).
  • Synthesis summary: entries added / rejected (with reason) / merged with existing.

Workflow

Step 1 — Read upstream artifacts

Read every listed artifact fully. Missing files are marked [MISSING] in the summary — do not fabricate entries for them.

Step 2 — Extract candidate entries (inclusion gate)

For each finding in each artifact, apply the gate. PASS requires ALL four:

  • Decision-relevant — affects future agent decisions (not a one-off fact of the current task)
  • Durable — still likely relevant in 90 days (the default expiry)
  • Non-duplicative — not self-evident from code, README, or AGENTS.md
  • Safe — not a secret, credential, or .env content

Any failure → REJECT, and record the reason in the summary (transparency, not silence). When in doubt, reject: the cost of a stale entry is paid by every future session that reads it.

Step 3 — Deduplicate and boundary-check

  1. Read existing ## Active entries of PROJECT-CONTEXT.md; absorb candidates that restate an existing fact (merge, don't duplicate).
  2. Boundary rules — content that belongs elsewhere is excluded here:
    • Rules, commands, conventions → AGENTS.md (S20)
    • Formal documentation → docs/ (S21)
    • Long-term decisions with rationale → decision records (e.g. decision-memory skill / ADRs)
    • This file holds informal but decision-relevant project knowledge: gotchas, half-finished state, historical constraints, warnings.
  3. For each surviving candidate assign: source (which artifact), scope (global | module/area name), expiry (default 90 days from today; never only for conventions that must not expire).

Step 4 — Batch confirm

Present ALL candidates at once as a table:

#Candidate entrySourceScopeExpiry

Ask once: approve all / adjust (specify rows) / reject (specify rows). Do NOT run project-context add per entry for bulk synthesis — its per-entry confirmation flow does not scale (measured in trial F1).

Step 5 — Append

Write approved entries into ## Active of PROJECT-CONTEXT.md (create the file with the standard header if absent). Follow the exact entry format from the project-context skill.

Step 6 — Verify (completion gate)

Before declaring done, all must hold:

  • Every entry has source, scope, and expiry
  • No duplicates with existing Active entries
  • No entry duplicates AGENTS.md / docs content
  • Active count within the 50-entry / ~200-line cap (warn if approaching; suggest project-context cleanup)

Anti-Patterns

  • Extracting everything — the gate exists to prevent context rot; a bloated file gets ignored.
  • Per-entry confirmation for bulk synthesis — batch confirm once, adjust by row.
  • Writing rules/commands — belongs in AGENTS.md, not the context file.
  • Fabricating entries for missing artifacts — mark [MISSING] and move on.
  • Skipping expiry — default 90 days; never requires explicit justification.
  • Duplicating an existing Active entry — merge or reject; duplicates halve the file's signal.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.